Multi-energy complementary hybrid energy storage scheduling system and scheduling method thereof

Abstract

The present invention relates to a multi-energy complementary hybrid energy storage scheduling system and a scheduling method thereof. The system uses a deep neural network model based on artificial intelligence technology to extract high-dimensional implicit features in the time dimension from the power consumption value of each electricity user, the valley power density of the power grid, and the energy storage capacity of the energy storage system, and uses this to perform feature fusion to judge the rationality of the proportion of electricity allocated to electricity users and the energy storage system. In addition, when fusion is performed, a class condition boundary constraint synthesis method of eigenvalues ​​is adopted to avoid excessive fragmentation of the decision area within the classification target domain due to out-of-distribution samples of the eigenvalue set, thereby improving the classification performance of the classification feature matrix and further improving the accuracy of rationality judgment.

Description

Technical Field

[0001] The present invention belongs to the technical field of energy storage intelligent control, and specifically relates to a multi-energy complementary hybrid energy storage scheduling system and a scheduling method thereof. Background Art

[0002] In recent years, the installed capacity of renewable energy sources such as wind power and photovoltaics has grown rapidly. However, due to the random nature of wind and solar power output and the dynamic fluctuations in user load demand, these renewable energy sources often exhibit inconsistencies and mismatches across time and space. Large-scale grid integration of renewable energy may pose new challenges to the safe and stable operation of the power system, potentially leading to curtailment of wind and solar power. Therefore, as the proportion of large-scale renewable energy connected to the grid continues to increase, how to absorb and absorb these renewable energy sources and ensure the safety and stability of the power system is an urgent international challenge that needs to be addressed.

[0003] Against the backdrop of rapidly increasing penetration of renewable energy, the regulatory pressure on the power grid continues to increase, and the demand for flexible resources continues to grow, necessitating the development of new regulatory resources to alleviate this pressure. Therefore, promoting the coupling and application of new large-capacity energy storage technologies with power generation and heating systems, and building new power generation systems such as distributed energy storage-flexible heating and power supply, and new Carnot batteries, can also be used to implement flexibility improvements on the coal-fired units that make up the largest portion of existing installed capacity. This will not only improve the structure of the power system, but also address the issue of absorbing the large-scale grid-connected renewable energy, alleviate system peak-shaving pressure, and ensure energy and power security.

[0004] When using off-peak electricity to charge an energy storage system, on the one hand, an appropriate proportion of off-peak electricity must be allocated to meet the electricity needs of electricity users. On the other hand, the energy storage system's energy storage efficiency must also be considered. It should be understood that electricity demand from electricity users fluctuates over time, that is, the overall electricity demand of electricity users will fluctuate. If this fluctuation is not taken into account, it may lead to fluctuations in user electricity consumption, that is, a decrease in power quality. Furthermore, when using off-peak electricity to charge an energy storage system, energy storage is a continuous process, while energy storage efficiency fluctuates as the state of the energy storage system changes. This change in energy storage efficiency can be represented by the remaining capacity to be stored.

[0005] Therefore, a technical solution is needed to determine the proportion of electricity distributed to electricity users and energy storage systems. Summary of the Invention

[0006] To overcome the shortcomings and deficiencies of the prior art, the present invention provides a multi-energy complementary hybrid energy storage scheduling system and a scheduling method thereof. The present invention can improve the classification performance of the classification feature matrix, thereby improving the accuracy of rationality judgment.

[0007] To achieve the above object, the technical solution adopted by the present invention is:

[0008] A multi-energy complementary hybrid energy storage scheduling system, comprising an electricity user data acquisition module, an electricity user data structuring module, an electricity consumption data encoding module, a transmission and storage data acquisition module, a time series encoding module, a responsiveness estimation module, a feature fusion module, a feature distribution correction module, and a scheduling result generation module;

[0009] An electricity user data collection module is used to obtain the power consumption value of each electricity user at multiple predetermined time points;

[0010] An electricity user data structuring module, configured to arrange the electricity consumption of each electricity user at a plurality of predetermined time points into a two-dimensional input matrix according to the electricity user sample dimension and the time dimension;

[0011] an electricity consumption data encoding module, configured to perform explicit spatial encoding on the two-dimensional input matrix using a first convolutional neural network model to obtain an electricity consumption feature matrix;

[0012] The power transmission and storage data acquisition module is used to obtain the off-peak power density of the power grid and the energy storage capacity of the energy storage system at multiple predetermined time points;

[0013] A time series encoding module is used to pass the valley power density of the power grid and the energy storage capacity of the energy storage system at multiple predetermined time points through a time series encoder including a one-dimensional convolutional layer to obtain a valley power density vector and an energy storage feature vector;

[0014] a responsiveness estimation module, configured to calculate a responsiveness matrix of a valley power density vector relative to the energy storage characteristic vector;

[0015] a feature fusion module, configured to fuse the responsiveness matrix and the power consumption feature matrix to obtain a classification feature matrix;

[0016] a feature distribution correction module, configured to perform eigenvalue-based class condition boundary constraint synthesis on the classification feature matrix to obtain a corrected classification feature matrix, wherein the eigenvalue-based class condition boundary constraint synthesis is performed based on the difference between a natural exponential function value raised to the power of the eigenvalue at each position in the classification feature matrix and the inverse of the eigenvalue at each position in the classification feature matrix;

[0017] The scheduling result generation module is used to pass the corrected classification feature matrix through the classifier to obtain a classification result, which is used to indicate whether the ratio of electric energy allocated to the energy storage system and allocated to electric users is reasonable.

[0018] Preferably, the electric user data structuring module includes a vector construction unit and a matrix construction unit;

[0019] a vector construction unit, configured to arrange the power consumption of each electricity user at a plurality of predetermined time points into row vectors according to the time dimension to obtain a plurality of row vectors;

[0020] The matrix construction unit is used to arrange multiple row vectors into a two-dimensional input matrix according to the dimension of the electric user sample.

[0021] Preferably, the electricity consumption data encoding module is further configured to use each layer of the first convolutional neural network model to perform the following operations on the input data in the forward pass of the layer:

[0022] Perform convolution on the input data to obtain a convolution feature map;

[0023] Performing mean pooling based on the local channel dimension on the convolutional feature map to obtain a pooled feature map; and

[0024] Performing nonlinear activation on the pooled feature map to obtain an activated feature map;

[0025] The output of the last layer of the first convolutional neural network is a power consumption feature matrix, and the input of the first layer of the first convolutional neural network is a two-dimensional input matrix.

[0026] Preferably, the temporal coding module includes an input vector construction unit, a fully connected coding unit and a one-dimensional convolutional coding unit;

[0027] An input vector construction unit, configured to arrange the valley power density of the power grid and the energy storage capacity of the energy storage system at a plurality of predetermined time points into one-dimensional input vectors according to the time dimension;

[0028] The fully connected encoding unit is used to use the fully connected layer of the temporal encoder to perform fully connected encoding on the input vector using the following formula to extract the high-dimensional implicit features of the eigenvalues ​​at each position in the input vector, wherein the formula is: Where X is the input vector, Y is the output vector, W is the weight matrix, B is the bias vector, Represents matrix multiplication;

[0029] The one-dimensional convolutional encoding unit is configured to perform one-dimensional convolutional encoding on the input vector using the one-dimensional convolutional layer of the temporal encoder according to the following formula to extract high-dimensional implicit correlation features between the eigenvalues ​​at each position in the input vector, wherein the formula is:

[0030]

[0031] Among them, a is the width of the convolution kernel in the x direction, F is the convolution kernel parameter vector, G is the local vector matrix operated with the convolution kernel function, and w is the size of the convolution kernel.

[0032] Preferably, the responsiveness estimation module is further configured to calculate a transfer matrix of a valley power density vector relative to an energy storage characteristic vector as the responsiveness matrix, wherein the transfer matrix multiplied by the valley power density vector is equal to the energy storage characteristic vector.

[0033] Preferably, the feature fusion module is further configured to: fuse the responsiveness matrix and the power consumption feature matrix using the following formula to obtain a classification feature matrix;

[0034] Wherein, the formula is:

[0035] M=αM1+βM2

[0036] Where M is the classification feature matrix, M1 is the responsiveness matrix, M2 is the electricity consumption feature matrix, and α and β are weighting parameters used to control the balance between the responsiveness matrix and the electricity consumption feature matrix.

[0037] Preferably, the feature distribution correction module is further configured to: perform a class condition boundary constraint synthesis based on eigenvalues ​​on the classification feature matrix using the following formula to obtain a corrected classification feature matrix;

[0038] Wherein, the formula is:

[0039]

[0040] Among them, m i，j Represents the eigenvalues ​​of each position in the classification feature matrix, m i，j′ Represents the eigenvalues ​​of each position in the corrected classification feature matrix.

[0041] Preferably, the scheduling result generating module is further configured to: the classifier processes the corrected classification feature matrix using the following formula to generate a classification result;

[0042] Wherein, the formula is:

[0043] softmax{(W n , B n ):...(W1,B1)|Project(F)}

[0044] Among them, Project(F) represents the projection of the corrected classification feature matrix into a vector, W1 to W n is the weight matrix of each fully connected layer, B1 to B n Represents the bias matrix of each fully connected layer.

[0045] The present invention also provides a scheduling method for the multi-energy complementary hybrid energy storage scheduling system, comprising the following steps:

[0046] (1) obtaining the power consumption values ​​of each electricity user at a plurality of predetermined time points;

[0047] (2) arranging the power consumption of each electricity user at a plurality of predetermined time points into a two-dimensional input matrix according to the electricity user sample dimension and the time dimension;

[0048] (3) Using the first convolutional neural network model to perform explicit spatial encoding on the two-dimensional input matrix to obtain the electricity consumption feature matrix;

[0049] (4) obtaining the valley power density of the power grid and the energy storage capacity of the energy storage system at multiple predetermined time points;

[0050] (5) The valley power density of the power grid and the energy storage capacity of the energy storage system at multiple predetermined time points are respectively passed through a temporal encoder including a one-dimensional convolutional layer to obtain a valley power density vector and an energy storage feature vector;

[0051] (6) Calculate the responsiveness matrix of the valley power density vector relative to the energy storage eigenvector;

[0052] (7) fusing the responsiveness matrix and the electricity consumption feature matrix to obtain a classification feature matrix;

[0053] (8) performing eigenvalue-based class condition boundary constraint synthesis on the classification feature matrix to obtain a corrected classification feature matrix, wherein the eigenvalue-based class condition boundary constraint synthesis is performed based on the difference between a natural exponential function value raised to a power of the eigenvalue of each position in the classification feature matrix and the inverse of the eigenvalue of each position in the classification feature matrix;

[0054] (9) The corrected classification feature matrix is ​​passed through a classifier to obtain a classification result. The classification result is used to indicate whether the ratio of electric energy allocated to the energy storage system and that allocated to the electric energy user is reasonable.

[0055] Preferably, in the scheduling method of the above-mentioned multi-energy complementary hybrid energy storage scheduling system, the electricity consumption of each electricity user at multiple predetermined time points is arranged into a two-dimensional input matrix according to the electricity user sample dimension and the time dimension, including: arranging the electricity consumption of each electricity user at multiple predetermined time points into row vectors according to the time dimension to obtain multiple row vectors; and arranging the multiple row vectors into a two-dimensional input matrix according to the electricity user sample dimension.

[0056] In the scheduling method of the above-mentioned multi-energy complementary hybrid energy storage scheduling system, the first convolutional neural network model is used to perform explicit spatial encoding on the two-dimensional input matrix to obtain a power consumption feature matrix, including: using each layer of the first convolutional neural network model to perform the following on the input data in the forward pass of the layer: convolution processing on the input data to obtain a convolution feature map; mean pooling based on the local channel dimension on the convolution feature map to obtain a pooled feature map; and nonlinear activation on the pooled feature map to obtain an activation feature map; wherein the output of the last layer of the first convolutional neural network is the power consumption feature matrix, and the input of the first layer of the first convolutional neural network is the two-dimensional input matrix.

[0057] In the above-mentioned scheduling method for the multi-energy complementary hybrid energy storage scheduling system, the valley power density of the power grid and the energy storage capacity of the energy storage system at the plurality of predetermined time points are respectively passed through a temporal encoder including a one-dimensional convolutional layer to obtain a valley power density vector and an energy storage feature vector, comprising: arranging the valley power density of the power grid and the energy storage capacity of the energy storage system at the plurality of predetermined time points into a one-dimensional input vector according to the time dimension; and using the fully connected layer of the temporal encoder to perform full connection encoding on the input vector using the following formula to extract the high-dimensional implicit features of the eigenvalues ​​at each position in the input vector, wherein the formula is: Where X is the input vector, Y is the output vector, W is the weight matrix, B is the bias vector, represents matrix multiplication; and, using the one-dimensional convolution layer of the temporal encoder to perform one-dimensional convolution encoding on the input vector using the following formula to extract high-dimensional implicit correlation features between the eigenvalues ​​at each position in the input vector, wherein the formula is:

[0058]

[0059] Among them, a is the width of the convolution kernel in the x direction, F is the convolution kernel parameter vector, G is the local vector matrix operated with the convolution kernel function, and w is the size of the convolution kernel.

[0060] In the scheduling method of the above-mentioned multi-energy complementary hybrid energy storage scheduling system, calculating the responsiveness matrix of the valley electricity power density vector relative to the energy storage characteristic vector includes: calculating the transfer matrix of the valley electricity power density vector relative to the energy storage characteristic vector as the responsiveness matrix, wherein the transfer matrix multiplied by the valley electricity power density vector is equal to the energy storage characteristic vector.

[0061] In the above-mentioned scheduling method of the multi-energy complementary hybrid energy storage scheduling system, fusing the responsiveness matrix and the power consumption characteristic matrix to obtain a classification characteristic matrix includes: fusing the responsiveness matrix and the power consumption characteristic matrix to obtain the classification characteristic matrix according to the following formula;

[0062] Wherein, the formula is:

[0063] M=αM1+βM2

[0064] Wherein, M is the classification feature matrix, M1 is the responsiveness matrix, M2 is the power consumption feature matrix, “+” represents the addition of the elements at corresponding positions of the responsiveness matrix and the power consumption feature matrix, and α and β are weighting parameters used to control the balance between the responsiveness matrix and the power consumption feature matrix.

[0065] In the above-mentioned scheduling method of the multi-energy complementary hybrid energy storage scheduling system, performing a class condition boundary constraint synthesis based on an eigenvalue on the classification feature matrix to obtain a corrected classification feature matrix, including: performing a class condition boundary constraint synthesis based on an eigenvalue on the classification feature matrix according to the following formula to obtain the corrected classification feature matrix;

[0066] Wherein, the formula is:

[0067]

[0068] Among them, m i，j Represents the eigenvalues ​​of each position in the classification feature matrix, m i，j′ Represents the eigenvalues ​​of each position in the corrected classification feature matrix.

[0069] In the above-mentioned scheduling method for the multi-energy complementary hybrid energy storage scheduling system, the corrected classification feature matrix is ​​passed through a classifier to obtain a classification result. The classification result is used to indicate whether the ratio of electric energy allocated to the energy storage system and to the electric energy user is reasonable. The method includes: the classifier processes the corrected classification feature matrix using the following formula to generate a classification result.

[0070] Wherein, the formula is: softmax{(W n , B n ):...:(W1, B1)|Project(F)}, where Project(F) represents the projection of the corrected classification feature matrix into a vector, W1 to W n is the weight matrix of each fully connected layer, B1 to B n Represents the bias matrix of each fully connected layer.

[0071] Compared with the prior art, the present invention has the following beneficial effects:

[0072] The present invention adopts a deep neural network model based on artificial intelligence technology to extract high-dimensional implicit features in the time dimension from the electricity power values ​​of each electricity user, the valley power density of the power grid, and the energy storage capacity of the energy storage system, and uses this to perform feature fusion to judge the rationality of the proportion of electricity allocated to electricity users and energy storage systems. In addition, when fusing features, a class condition boundary constraint synthesis method of eigenvalues ​​is adopted to avoid excessive fragmentation of the decision area within the classification target domain due to out-of-distribution samples of the eigenvalue set, thereby improving the classification performance of the classification feature matrix and further improving the accuracy of rationality judgment. BRIEF DESCRIPTION OF THE DRAWINGS

[0073] Figure 1 A roadmap for energy storage thermoelectric technology in a multi-energy complementary hybrid energy storage scheduling system according to an embodiment of the present invention;

[0074] Figure 2 Schematic diagram of the energy storage process during off-peak period in a multi-energy complementary hybrid energy storage scheduling system according to an embodiment of the present invention;

[0075] Figure 3 Schematic diagram of the energy release process during peak / peak periods in a multi-energy complementary hybrid energy storage scheduling system according to an embodiment of the present invention;

[0076] Figure 4 Schematic diagram of the process of the molten salt heat storage system in the multi-energy complementary hybrid energy storage scheduling system according to an embodiment of the present invention;

[0077] Figure 5 A structural diagram of a molten salt tank and its insulation infrastructure in a multi-energy complementary hybrid energy storage scheduling system according to an embodiment of the present invention;

[0078] Figure 6 Schematic diagram of the power storage system flow in a multi-energy complementary hybrid energy storage scheduling system according to an embodiment of the present invention;

[0079] Figure 7 This is an application scenario diagram of a multi-energy complementary hybrid energy storage scheduling system according to an embodiment of the present invention;

[0080] Figure 8 A block diagram of a multi-energy complementary hybrid energy storage scheduling system according to an embodiment of the present invention;

[0081] Figure 9 Flowchart of a scheduling method for a multi-energy complementary hybrid energy storage scheduling system according to an embodiment of the present invention;

[0082] Figure 10 Schematic diagram of the architecture of a scheduling method for a multi-energy complementary hybrid energy storage scheduling system according to an embodiment of the present invention. DETAILED DESCRIPTION

[0083] The technical solutions of the present invention are further specifically described below through examples. These examples are provided for the purpose of illustrating the present invention and are not intended to limit the present invention. All other examples obtained by persons of ordinary skill in the art based on the examples of the present invention without creative effort are intended to fall within the scope of protection of the present invention.

[0084] Against the backdrop of rapidly increasing penetration of renewable energy, the regulatory pressure on the power grid continues to increase, and the demand for flexible resources continues to grow, necessitating the development of new regulatory resources to alleviate this pressure. Therefore, promoting the coupling and application of new large-capacity energy storage technologies with power generation and heating systems, and building new power generation systems such as distributed energy storage-flexible heating and power supply, and new Carnot batteries, can also be used to implement flexibility improvements on the coal-fired units that make up the largest portion of existing installed capacity. This will not only improve the structure of the power system, but also address the issue of absorbing the large-scale grid-connected renewable energy, alleviate system peak-shaving pressure, and ensure energy and power security.

[0085] To meet the construction needs of green, low-carbon industrial parks, this paper proposes an energy storage solution that combines molten salt heat storage and electricity storage technologies, taking into account the applicability and affordability of various energy storage technologies. This solution establishes a multi-energy, low-carbon, hybrid energy storage and flexible peak-shaving thermal power system. The system includes a molten salt heat storage system, an electricity storage system, and a power generation and heating system.

[0086] The energy storage system is powered by off-peak electricity and various renewable energy sources (medium-pressure steam serves as a backup source). During off-peak periods, off-peak electricity is used for storage and to generate low-pressure steam for direct heating users. During peak and peak periods, thermal energy is released to generate electricity and steam, while also releasing electricity, addressing peak energy consumption bottlenecks for heat and electricity users in the park. The energy storage system can also accommodate large-scale renewable energy sources such as wind and solar power, as well as flexibly regulate the peak load of existing units. It can serve as a mobile energy storage power source.

[0087] Energy storage thermoelectric technology roadmap Figure 1 shown.

[0088] Energy storage process: During off-peak hours, or after large-scale wind and solar power are connected, valley electricity is used to heat the cold molten salt to the required temperature; valley electricity is used to directly heat desalted water to generate low-pressure steam for heating users; valley electricity is used to charge flow batteries and lithium iron phosphate batteries for energy storage.

[0089] Energy release process: During peak hours and peak periods, or when the output of new energy sources such as wind and solar power cannot meet the demand for electricity and heat, high-temperature molten salt is used to heat desalted water to generate medium-pressure steam, which enters the back-pressure steam turbine to generate electricity, and the exhaust steam and electricity are provided to users; the energy stored in the flow battery and lithium iron phosphate battery is released to supply electricity to users, and can also be used as a mobile power supply.

[0090] The process of storing energy during low-peak periods and releasing energy during high-peak periods is as follows: Figure 2 and 3 shown.

[0091] 1. Molten salt heat storage system

[0092] a.Technical route

[0093] The molten salt heat storage system includes a molten salt heat storage module, a molten salt electric heating module, a molten salt heat release module, a molten salt steam heating module, and a desalted water electric heating module.

[0094] Molten salt heat storage: Utilizing off-peak electricity as a power source, a molten salt pump pumps low-temperature molten salt to the molten salt electric heating module during off-peak periods. The molten salt electric heating module uses off-peak electricity to heat the low-temperature molten salt to the desired temperature using an electric heater. The heat is then transferred to the molten salt heat storage module, which stores the high-temperature molten salt in a high-temperature molten salt tank. Simultaneously, the desalted water electric heating module uses off-peak electricity to directly heat the desalted water, generating low-pressure steam for heat users. Both the molten salt and desalted water electric heating modules are shut down after off-peak periods.

[0095] Molten salt heat release: A molten salt pump is used to pump high-temperature molten salt to the molten salt heat release module. During peak hours and peak periods, or when the output of new energy sources such as wind and solar power cannot meet the electricity and heat demand, the high-temperature molten salt is used to heat the desalted water and generate medium-pressure steam, which enters the back-pressure steam turbine to generate electricity. The exhaust steam and electricity are provided to users, and the high-temperature molten salt is returned to the low-temperature molten salt tank after heat exchange.

[0096] The molten salt heat storage system process is as follows Figure 4 shown.

[0097] b. Key equipment technology

[0098] The key equipment and processes for the operation of the molten salt heat storage system include molten salt, molten salt storage tanks, electric heaters, heat exchangers, molten salt pumps, salt drain tanks, steam turbines, generators and pipeline valves.

[0099] 1) Molten salt

[0100] Molten salt can be divided into single salt, binary salt, ternary salt and multi-component salt according to the number of components. Nitrate is preferred in this system.

[0101] 2) Molten salt storage tank

[0102] Including high-temperature storage tanks and low-temperature storage tanks, the main places for molten salt storage.

[0103] The volume of the storage tank is determined based on the heat storage capacity and the physical properties of the molten salt. Multiple temperature measuring points are set on the tank wall to monitor the molten salt temperature at any time.

[0104] The foundation of a molten salt storage tank utilizes a load-bearing and thermally insulating structure. While the foundation provides support, the insulation prevents the heat stored in the tank from escaping, minimizing energy loss. The foundation must meet the tank's load-bearing, thermal insulation, and expansion requirements. The tank bottom foundation primarily consists of a sand cushion, steel plate, insulating refractory bricks, foam glass insulation, and high-temperature concrete.

[0105] The structure of the molten salt tank insulation foundation is as follows Figure 5 As shown, the structure of the molten salt tank insulation foundation includes: a tank body 1; a sand cushion layer 2; a steel plate layer 3; a thermal insulation layer 4; an annular thermal insulation layer 5; a high-temperature resistant concrete outer wall 6; a ventilation pipe 7; a heat dissipation layer 8; a refractory concrete layer 9; a refractory brick inner wall 10; and a concrete support layer 11.

[0106] A safety valve is installed at the connection between the molten salt storage tank and the molten salt pump, and the design pressure is atmospheric pressure. Currently, there is no design standard in the thermal energy storage industry. The design of molten salt storage tanks refers to the design standard for petroleum storage tanks, API650.

[0107] 3) Electric heater

[0108] Includes electric pipe heaters and anti-condensation electric heaters.

[0109] Through electric heaters, during off-peak periods, low-temperature molten salt can be heated into high-temperature molten salt using off-peak electricity, and the use requirements can be met through molten salt heat storage; at the same time, desalted water can be heated into low-pressure steam using off-peak electricity for use by nighttime heat users.

[0110] The anti-condensation electric heater is installed at the bottom of the molten salt storage tank to compensate for the heat loss of the tank wall and the tank bottom. It can ensure that the molten salt temperature is maintained at not lower than the melting point when the system stops running. At the same time, a self-measuring system and a control system are set. When the molten salt temperature is lower than the set minimum temperature, it starts to heat the molten salt to the set temperature.

[0111] 4) Heat exchanger

[0112] The heat exchanger type adopts shell and tube heat exchanger according to the heat exchanger design manual and heating hot water requirements.

[0113] The heat exchanger used in the molten salt heat release module includes a preheater, evaporator, and superheater. The preheater heats the desalted water to a temperature slightly below the critical temperature for evaporation. The evaporator converts the preheated desalted water into slightly superheated steam. The superheater heats the slightly superheated steam to the required temperature. The heat exchanger uses a U-tube surface structure. The molten salt in the preheater and superheater flows through the shell, while the steam in the evaporator flows through the shell.

[0114] During installation, the heat exchanger's manufacturing, outlet temperature, outlet pressure and other related indicators are tested to prove that its performance meets the requirements of the adopted specifications and standards, meets the requirements of the 30-year design life, and ensures that the product quality meets the requirements of the standards and specifications.

[0115] 5) Molten salt pump

[0116] The molten salt pumps utilize long-axis submersible pumps, with a high-temperature molten salt pump and a low-temperature molten salt pump installed on the high-temperature storage tank and low-temperature molten salt pump, respectively. Driven by variable-frequency motors with high reliability and low maintenance requirements, the molten salt circulates between the high-temperature storage tank, heating system, heat release system, and low-temperature storage tank. By using variable frequency drives (VFDs), the flow rate within the energy storage system can be adapted to varying heating conditions, maximizing energy storage efficiency.

[0117] 6) Salt tank

[0118] The salt drain tank is primarily used to collect molten salt from pipelines and heat exchangers. A vertical pump returns the collected molten salt to the cold salt tank, and an electric heater is installed to prevent condensation. Leak detectors are installed in the salt drain system. If a leak is detected, the salt / water pipelines are immediately separated and the leak location is confirmed.

[0119] 7) Steam turbine and generator

[0120] The high-temperature steam generated by the heat release of the molten salt enters the steam turbine and drives the generator to generate electricity, and the exhaust steam and electricity are provided to users.

[0121] 8) Pipeline valves

[0122] The main channel of the molten salt transportation process needs to be heated and preheated to ensure that the molten salt remains in liquid state under working and non-working conditions. An electric heating system is used to prevent freezing and blockage.

[0123] Pipelines require thermal insulation to ensure they maintain a preheated temperature during operation and to minimize heat loss, lower consumption, and save production costs. Generally, during normal production, the maximum insulation wall temperature should not exceed 50°C above room temperature.

[0124] 2. Power storage system

[0125] a.Technical route

[0126] The energy storage system includes flow batteries and lithium iron phosphate batteries.

[0127] Battery charging: Valley electricity is used as the power source. During low-voltage periods, valley electricity is used to charge liquid flow batteries and lithium iron phosphate batteries.

[0128] Battery discharge: During peak hours and peak periods of electricity consumption, battery discharge not only provides the electricity required for self-operation, but the remaining electricity is used by electricity users in the park. It can be used as a fixed power source or as a mobile power source.

[0129] The process of the power storage system is as follows Figure 6 shown.

[0130] b. Key equipment technology

[0131] The key equipment and processes of the power storage system include energy storage batteries, PCS energy storage converters, BMS battery management systems, EMS energy management systems, etc.

[0132] It should be understood that when using valley electricity to charge an energy storage system, on the one hand, the valley electricity needs to allocate an appropriate proportion of electricity to meet the electricity needs of electricity users, and on the other hand, the energy storage efficiency of the energy storage system also needs to be considered. It should be understood that the electricity demand of electricity users fluctuates in the time series dimension. In other words, the overall electricity demand of electricity users will fluctuate. If this fluctuation factor is not taken into account, it may cause fluctuations in user electricity consumption, that is, cause a decline in power quality. Moreover, when using valley electricity to charge an energy storage system, energy storage is a continuous process, while energy storage efficiency will fluctuate as the state of the energy storage system changes. The change in energy storage efficiency can be represented by the remaining capacity to be stored. Therefore, a technical solution is needed to determine the proportion of electricity allocated to electricity users and the energy storage system.

[0133] Accordingly, in the technical solution of the present invention, the electricity consumption of each electricity user at each predetermined time point is selected to represent its electricity demand. That is, specifically, first obtain the electricity consumption value of each electricity user at a plurality of predetermined time points, and arrange the electricity consumption of each electricity user at the plurality of predetermined time points into a two-dimensional input matrix according to the electricity user sample dimension and the time dimension. In this way, the data distribution information of the electricity consumption of each electricity user at the plurality of predetermined time points in the sample and time dimensions can be integrated. Then, a first convolutional neural network model with excellent performance in implicit correlation feature extraction is used to perform explicit spatial encoding on the two-dimensional input matrix to extract the high-dimensional implicit correlation feature information of the electricity consumption of each electricity user at each predetermined time point, thereby obtaining an electricity consumption feature matrix.

[0134] It should be understood that when using valley power to charge an energy storage system, energy storage is a continuous process, but the energy storage efficiency will fluctuate as the state of the energy storage system changes, and the change in energy storage efficiency can be represented by the remaining capacity to be stored. Therefore, further, the valley power density of the power grid and the energy storage capacity of the energy storage system at the multiple predetermined time points are obtained and encoded in a time series encoder including a one-dimensional convolutional layer to mine the dynamic change characteristics of the valley power density of the power grid and the energy storage capacity of the energy storage system in the time series dimension, thereby obtaining the valley power density vector and the energy storage feature vector.

[0135] Then, considering that the valley power density characteristic of the power grid is a responsive characteristic of the energy storage capacity characteristic of the energy storage system, in order to strengthen the correlation characteristic information between the two and obtain dynamic implicit change characteristics, the transfer matrix of the valley power density vector relative to the energy storage characteristic vector is further calculated as the responsiveness matrix. In this way, the responsiveness matrix and the power consumption characteristic matrix can be fused to obtain a classification characteristic matrix. For example, the classification characteristic matrix can be obtained by calculating the weighted sum of the responsiveness matrix and the power consumption characteristic matrix according to position.

[0136] It should be understood that since the responsiveness matrix corresponds to the time series-time series dimension, and the power consumption feature matrix corresponds to the time series-sample dimension, when they are fused, there may be out-of-distribution eigenvalues ​​due to dimensional misalignment. Therefore, for the classification feature matrix obtained by fusing the responsiveness matrix and the power consumption feature matrix, a class condition boundary constraint synthesis of the eigenvalues ​​is performed, which is expressed as:

[0137]

[0138] This class-condition boundary-constrained synthesis of eigenvalues ​​utilizes a structured, rule-based understanding of the eigenvalues ​​and their class conditions to perform boundary-constrained synthesis of features. This prevents excessive fragmentation of the decision region within the classification target domain caused by out-of-distribution samples in the eigenvalue set, improving the classification performance of the classification feature matrix. This, in turn, results in a more rational ratio of energy allocated to the energy storage system and to electricity users.

[0139] Based on this, the present invention proposes a multi-energy complementary hybrid energy storage scheduling system, which includes: an electricity user data acquisition module for obtaining the electricity power value of each electricity user at multiple predetermined time points; an electricity user data structuring module for arranging the electricity power of each electricity user at the multiple predetermined time points into a two-dimensional input matrix according to the electricity user sample dimension and the time dimension; an electricity data encoding module for using a first convolutional neural network model to perform explicit spatial encoding on the two-dimensional input matrix to obtain an electricity feature matrix; a power transmission and storage data acquisition module for obtaining the valley power density of the power grid and the energy storage capacity of the energy storage system at the multiple predetermined time points; a time series encoding module for respectively passing the valley power density of the power grid and the energy storage capacity of the energy storage system at the multiple predetermined time points through a time series encoder including a one-dimensional convolution layer to obtain the valley power density vector and the storage capacity vector. energy characteristic vector; a responsiveness estimation module for calculating the responsiveness matrix of the valley power power density vector relative to the energy storage characteristic vector; a feature fusion module for fusing the responsiveness matrix and the power consumption characteristic matrix to obtain a classification characteristic matrix; a feature distribution correction module for performing a class condition boundary constraint synthesis based on eigenvalues ​​on the classification characteristic matrix to obtain a corrected classification characteristic matrix, wherein the class condition boundary constraint synthesis based on eigenvalues ​​is based on the difference between the value of a natural exponential function raised to the power of the eigenvalues ​​of each position in the classification characteristic matrix and the inverse of the eigenvalues ​​of each position in the classification characteristic matrix; and a scheduling result generation module for passing the corrected classification characteristic matrix through a classifier to obtain a classification result, wherein the classification result is used to indicate whether the proportion of electric energy allocated to the energy storage system and the proportion of electric energy allocated to the electric user is reasonable.

[0140] The application scenario of the multi-energy complementary hybrid energy storage scheduling system according to an embodiment of the present invention is shown in the figure below: Figure 7 As shown, in this application scenario, first, by deploying at each electricity user's electricity usage site (for example, Figure 7 Each power measuring instrument (e.g., Figure 7 The power consumption values ​​of each electricity user at a plurality of predetermined time points are obtained by using various sensors (e.g., Figure 7 Then, the power consumption values ​​of each electricity user at the plurality of predetermined time points and the power grid (for example, Figure 7 The valley power density of the energy storage system (e.g., Figure 7 The energy storage capacity of E) shown in FIG is input into a server that deploys a multi-energy complementary hybrid energy storage scheduling algorithm (for example, Figure 7The cloud server S shown in the figure) is capable of processing the power consumption values ​​of each electricity user at the multiple predetermined time points, the valley power density of the power grid, and the energy storage capacity of the energy storage system using a multi-energy complementary hybrid energy storage scheduling algorithm to generate a classification result indicating whether the ratio of electricity allocated to the energy storage system and to the electricity user is reasonable.

[0141] After introducing the basic principles of the present invention, various non-limiting embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0142] Example 1

[0143] The block diagram of the multi-energy complementary hybrid energy storage scheduling system according to an embodiment of the present invention is as follows: Figure 8 As shown, according to an embodiment of the present invention, a multi-energy complementary hybrid energy storage scheduling system 200 includes: an electricity user data acquisition module 210, which is used to obtain the electricity power value of each electricity user at a plurality of predetermined time points; an electricity user data structuring module 220, which is used to arrange the electricity power of each electricity user at the plurality of predetermined time points into a two-dimensional input matrix according to the electricity user sample dimension and the time dimension; an electricity data encoding module 230, which is used to perform explicit spatial encoding on the two-dimensional input matrix using a first convolutional neural network model to obtain an electricity feature matrix; a power transmission and storage data acquisition module 240, which is used to obtain the valley power density of the power grid and the energy storage capacity of the energy storage system at the plurality of predetermined time points; a time series encoding module 250, which is used to respectively pass the valley power density of the power grid and the energy storage capacity of the energy storage system at the plurality of predetermined time points through a time series encoder including a one-dimensional convolution layer to obtain the valley power density vector. quantity and energy storage characteristic vector; a responsiveness estimation module 260, for calculating the responsiveness matrix of the valley power density vector relative to the energy storage characteristic vector; a feature fusion module 270, for fusing the responsiveness matrix and the power consumption characteristic matrix to obtain a classification feature matrix; a feature distribution correction module 280, for performing a class condition boundary constraint synthesis based on eigenvalues ​​on the classification feature matrix to obtain a corrected classification feature matrix, wherein the class condition boundary constraint synthesis based on eigenvalues ​​is based on the difference between the value of a natural exponential function raised to the power of the eigenvalues ​​of each position in the classification feature matrix and the inverse of the eigenvalues ​​of each position in the classification feature matrix; and a scheduling result generation module 290, for passing the corrected classification feature matrix through a classifier to obtain a classification result, wherein the classification result is used to indicate whether the proportion of electric energy allocated to the energy storage system and allocated to the electric user is reasonable.

[0144] Specifically, in an embodiment of the present invention, the electricity user data acquisition module 210, the electricity user data structuring module 220, and the electricity consumption data encoding module 230 are used to obtain the electricity power values ​​of each electricity user at multiple predetermined time points, and arrange the electricity power of each electricity user at the multiple predetermined time points into a two-dimensional input matrix according to the electricity user sample dimension and the time dimension, and then use the first convolutional neural network model to perform explicit spatial encoding on the two-dimensional input matrix to obtain an electricity consumption feature matrix. As mentioned above, when using valley electricity to charge the energy storage system, on the one hand, valley electricity needs to allocate an appropriate proportion of electricity to meet the electricity demand of the electricity user, and on the other hand, the energy storage efficiency of the energy storage system also needs to be considered. It should be understood that the electricity demand of the electricity user fluctuates in the time series dimension, that is, the overall electricity demand of the electricity user will fluctuate. If the fluctuation factor is not taken into account, it may cause the user's electricity consumption to fluctuate, that is, cause a decline in power quality. Furthermore, when using valley electricity to charge an energy storage system, energy storage is a continuous process, but energy storage efficiency fluctuates as the energy storage system's state changes. This change in energy storage efficiency can be represented by the remaining capacity to be stored. Therefore, in the technical solution of the present invention, it is desirable to represent the electricity demand of each electricity user by the power consumption at each predetermined time point.

[0145] That is, specifically, first obtain the power consumption values ​​of each electricity user at a plurality of predetermined time points, and arrange the power consumption of each electricity user at the plurality of predetermined time points into a two-dimensional input matrix according to the electricity user sample dimension and the time dimension. In this way, the data distribution information of the power consumption of each electricity user at the plurality of predetermined time points in the sample and time dimensions can be integrated. Accordingly, in a specific example, the power consumption of each electricity user at the plurality of predetermined time points is arranged into row vectors according to the time dimension to obtain a plurality of row vectors; and the plurality of row vectors are arranged into the two-dimensional input matrix according to the electricity user sample dimension. Then, the first convolutional neural network model with excellent performance in implicit correlation feature extraction is used to perform explicit spatial encoding on the two-dimensional input matrix to extract high-dimensional implicit correlation feature information of the power consumption of each electricity user at each predetermined time point, thereby obtaining an electricity consumption feature matrix.

[0146] More specifically, in an embodiment of the present invention, the electricity consumption data encoding module is further used to: use each layer of the first convolutional neural network model to perform the following on the input data in the forward pass of the layer: convolution processing on the input data to obtain a convolution feature map; mean pooling based on the local channel dimension on the convolution feature map to obtain a pooled feature map; and nonlinear activation on the pooled feature map to obtain an activation feature map; wherein the output of the last layer of the first convolutional neural network is the electricity consumption feature matrix, and the input of the first layer of the first convolutional neural network is the two-dimensional input matrix.

[0147] Specifically, in an embodiment of the present invention, the power transmission and storage data acquisition module 240 and the time series encoding module 250 are used to obtain the valley power density of the power grid and the energy storage capacity of the energy storage system at the multiple predetermined time points, and pass the valley power density of the power grid and the energy storage capacity of the energy storage system at the multiple predetermined time points through a time series encoder including a one-dimensional convolution layer to obtain a valley power density vector and an energy storage feature vector. It should be understood that when charging the energy storage system using valley power, energy storage is a continuous process, but the energy storage efficiency will fluctuate as the state of the energy storage system changes, and the change in energy storage efficiency can be represented by the remaining capacity to be stored. Therefore, in the technical solution of the present invention, further, the valley power density of the power grid and the energy storage capacity of the energy storage system at the multiple predetermined time points are obtained, and are respectively encoded in a time series encoder including a one-dimensional convolution layer to mine the dynamic change characteristics of the valley power density of the power grid and the energy storage capacity of the energy storage system in the time series dimension, thereby obtaining the valley power density vector and the energy storage feature vector.

[0148] More specifically, in an embodiment of the present invention, the time series encoding module includes: an input vector construction unit, configured to arrange the valley power density of the power grid and the energy storage capacity of the energy storage system at the plurality of predetermined time points into one-dimensional input vectors according to the time dimension; and a fully connected encoding unit, configured to use the fully connected layer of the time series encoder to perform fully connected encoding on the input vector using the following formula to extract high-dimensional implicit features of the eigenvalues ​​at each position in the input vector, wherein the formula is: Where X is the input vector, Y is the output vector, W is the weight matrix, and B is the bias vector, represents matrix multiplication; and a one-dimensional convolutional encoding unit, configured to perform one-dimensional convolutional encoding on the input vector using the one-dimensional convolutional layer of the temporal encoder according to the following formula to extract high-dimensional implicit correlation features between eigenvalues ​​at each position in the input vector, wherein the formula is:

[0149]

[0150] Among them, a is the width of the convolution kernel in the x direction, F is the convolution kernel parameter vector, G is the local vector matrix operated with the convolution kernel function, and w is the size of the convolution kernel.

[0151] Specifically, in an embodiment of the present invention, the responsiveness estimation module 260 and the feature fusion module 270 are used to calculate the responsiveness matrix of the valley power density vector relative to the energy storage feature vector, and fuse the responsiveness matrix and the power consumption feature matrix to obtain a classification feature matrix. It should be understood that, considering that the valley power density feature of the power grid is a responsive feature of the energy storage capacity feature of the energy storage system, in order to strengthen the correlation feature information between the two to obtain a dynamic implicit change feature, in the technical solution of the present invention, the transfer matrix of the valley power density vector relative to the energy storage feature vector is further calculated as the responsiveness matrix. Here, in a specific example, the transfer matrix multiplied by the valley power density vector is equal to the energy storage feature vector. In this way, the responsiveness matrix and the power consumption feature matrix can be further fused to obtain a classification feature matrix. For example, the weighted sum of the responsiveness matrix and the power consumption feature matrix by position can be calculated to obtain the classification feature matrix.

[0152] More specifically, in an embodiment of the present invention, the feature fusion module is further configured to: fuse the responsiveness matrix and the power consumption feature matrix using the following formula to obtain the classification feature matrix;

[0153] Wherein, the formula is:

[0154] M=αM1+βM2

[0155] Wherein, M is the classification feature matrix, M1 is the responsiveness matrix, M2 is the power consumption feature matrix, “+” represents the addition of the elements at corresponding positions of the responsiveness matrix and the power consumption feature matrix, and α and β are weighting parameters used to control the balance between the responsiveness matrix and the power consumption feature matrix.

[0156] Specifically, in an embodiment of the present invention, the feature distribution correction module 280 is used to perform a class condition boundary constraint synthesis based on eigenvalues ​​on the classification feature matrix to obtain a corrected classification feature matrix, wherein the class condition boundary constraint synthesis based on eigenvalues ​​is based on the difference between the natural exponential function value raised to the power of the eigenvalues ​​of each position in the classification feature matrix and the inverse of the eigenvalues ​​of each position in the classification feature matrix. It should be understood that since the responsiveness matrix corresponds to the time series-time series dimension, and the power consumption feature matrix corresponds to the time series-sample dimension, when they are fused, there may be out-of-distribution eigenvalues ​​due to dimensional misalignment. Therefore, in the technical solution of the present invention, for the classification feature matrix obtained by fusing the responsiveness matrix and the power consumption feature matrix, a class condition boundary constraint synthesis of eigenvalues ​​is performed. It should be understood that the class-condition boundary-constrained synthesis of eigenvalues ​​utilizes a structured, rule-based understanding of the eigenvalues ​​and their associated class conditions to perform boundary-constrained synthesis of features. This prevents excessive fragmentation of the decision region within the classification target domain caused by out-of-distribution samples in the eigenvalue set, thereby improving the classification performance of the classification feature matrix. Furthermore, this results in a more reasonable ratio of energy allocated to the energy storage system and to electricity users.

[0157] More specifically, in an embodiment of the present invention, the feature distribution correction module is further configured to: perform a class condition boundary constraint synthesis based on eigenvalues ​​on the classification feature matrix using the following formula to obtain the corrected classification feature matrix;

[0158] Wherein, the formula is:

[0159]

[0160] Among them, m i，j Represents the eigenvalues ​​of each position in the classification feature matrix, m i，j′ Represents the eigenvalues ​​of each position in the corrected classification feature matrix.

[0161] Specifically, in an embodiment of the present invention, the scheduling result generation module 290 is used to pass the corrected classification feature matrix through a classifier to obtain a classification result, and the classification result is used to indicate whether the ratio of electric energy allocated to the energy storage system and the electric energy allocated to the electric energy user is reasonable. That is, in a specific example, the classifier processes the corrected classification feature matrix according to the following formula to generate a classification result, wherein the formula is: softmax{(W n , B n):...:(W1, B1)|Project(F)}, where Project(F) represents the projection of the corrected classification feature matrix into a vector, W1 to W n is the weight matrix of each fully connected layer, B1 to B n Represents the bias matrix of each fully connected layer.

[0162] In summary, the multi-energy complementary hybrid energy storage scheduling system 200 based on the embodiment of the present invention is explained, which uses a deep neural network model of artificial intelligence technology to extract high-dimensional implicit features in the time dimension for the electricity power value of each electricity user, the valley power density of the power grid, and the energy storage capacity of the energy storage system, and uses this to perform feature fusion to judge the rationality of the proportion of electricity allocated to electricity users and energy storage systems. In addition, when fusing features, the class condition boundary constraint synthesis method of the eigenvalue is adopted to avoid excessive fragmentation of the decision area in the classification target domain due to the out-of-distribution samples of the eigenvalue set, thereby improving the classification performance of the classification feature matrix and thus improving the accuracy of the rationality judgment.

[0163] As described above, the multi-energy hybrid energy storage scheduling system 200 according to an embodiment of the present invention can be implemented in various terminal devices, such as a server for a multi-energy hybrid energy storage scheduling algorithm. In one example, the multi-energy hybrid energy storage scheduling system 200 according to an embodiment of the present invention can be integrated into a terminal device as a software module and / or a hardware module. For example, the multi-energy hybrid energy storage scheduling system 200 can be a software module in the operating system of the terminal device, or it can be an application developed specifically for the terminal device. Of course, the multi-energy hybrid energy storage scheduling system 200 can also be one of the many hardware modules of the terminal device.

[0164] Alternatively, in another example, the multi-energy complementary hybrid energy storage scheduling system 200 and the terminal device may also be separate devices, and the multi-energy complementary hybrid energy storage scheduling system 200 may be connected to the terminal device via a wired and / or wireless network and transmit interactive information in accordance with an agreed data format.

[0165] Example 2

[0166] A flow chart of a scheduling method for a multi-energy complementary hybrid energy storage scheduling system is shown in the figure. Figure 9As shown, the scheduling method of the multi-energy complementary hybrid energy storage scheduling system according to an embodiment of the present invention includes the following steps: S110, obtaining the power consumption value of each electricity user at a plurality of predetermined time points; S120, arranging the power consumption of each electricity user at the plurality of predetermined time points into a two-dimensional input matrix according to the electricity user sample dimension and the time dimension; S130, using a first convolutional neural network model to perform explicit spatial encoding on the two-dimensional input matrix to obtain a power consumption feature matrix; S140, obtaining the valley power density of the power grid and the energy storage capacity of the energy storage system at the plurality of predetermined time points; S150, respectively passing the valley power density of the power grid and the energy storage capacity of the energy storage system at the plurality of predetermined time points through a temporal encoder including a one-dimensional convolution layer to obtain a valley power density vector and an energy storage feature matrix. vector; S160, calculating the responsiveness matrix of the valley power density vector relative to the energy storage characteristic vector; S170, fusing the responsiveness matrix and the electricity consumption characteristic matrix to obtain a classification characteristic matrix; S180, performing a class condition boundary constraint synthesis based on eigenvalues ​​on the classification characteristic matrix to obtain a corrected classification characteristic matrix, wherein the class condition boundary constraint synthesis based on eigenvalues ​​is based on the difference between the value of a natural exponential function raised to the power of the eigenvalues ​​of each position in the classification characteristic matrix and the inverse of the eigenvalues ​​of each position in the classification characteristic matrix; and, S190, passing the corrected classification characteristic matrix through a classifier to obtain a classification result, wherein the classification result is used to indicate whether the proportion of electric energy allocated to the energy storage system and that allocated to the electricity user is reasonable.

[0167] The schematic diagram of the architecture of the scheduling method of the multi-energy complementary hybrid energy storage scheduling system according to an embodiment of the present invention is as follows: Figure 10 As shown, in the network architecture of the scheduling method of the multi-energy complementary hybrid energy storage scheduling system, first, the power consumption of each electricity user at the plurality of predetermined time points (for example, Figure 10 The P1 shown in FIG is arranged into a two-dimensional input matrix according to the electric user sample dimension and the time dimension (for example, Figure 10 Then, using the first convolutional neural network model (e.g., Figure 10 The CNN1 shown in FIG1 performs explicit spatial encoding on the two-dimensional input matrix to obtain the power consumption feature matrix (e.g., Figure 10 Then, the valley power density of the grid at the plurality of predetermined time points (for example, Figure 10 Q1 as shown in the figure) and the energy storage capacity of the energy storage system (e.g. Figure 10 Q2 as shown in FIG10 ) are respectively passed through a temporal encoder including a one-dimensional convolutional layer (eg, as shown in FIG10 E) to obtain a valley power density vector (eg, as shown in FIG10 ). Figure 10 VF1) and the energy storage characteristic vector (e.g., Figure 10 Then, the responsiveness matrix of the valley power density vector relative to the energy storage characteristic vector is calculated (for example, as Figure 10 Then, the responsiveness matrix and the power consumption feature matrix are fused to obtain a classification feature matrix (e.g., Figure 10 MF shown in FIG); Then, the classification feature matrix is ​​synthesized based on the class condition boundary constraint of the eigenvalue to obtain the corrected classification feature matrix (for example, Figure 10 MC shown in ); and, finally, passing the corrected classification feature matrix through a classifier (e.g., Figure 10 The classification result is used to indicate whether the ratio of electric energy allocated to the energy storage system and electric energy allocated to the electric user is reasonable.

[0168] More specifically, in steps S110, S120, and S130, the power consumption values ​​of each electricity user at multiple predetermined time points are obtained, and the power consumption values ​​of each electricity user at these predetermined time points are arranged into a two-dimensional input matrix according to the electricity user sample dimension and the time dimension. The two-dimensional input matrix is ​​then explicitly spatially encoded using a first convolutional neural network model to obtain a power consumption feature matrix. It should be understood that when using off-peak electricity to charge an energy storage system, on the one hand, an appropriate proportion of off-peak electricity must be allocated to meet the electricity demand of electricity users, while on the other hand, the energy storage system's energy storage efficiency must also be considered. It should be understood that electricity demand from electricity users fluctuates in the time series dimension, that is, the overall electricity demand of electricity users will fluctuate. If this fluctuation factor is not taken into account, it may lead to fluctuations in electricity consumption for users, i.e., a decrease in power quality. Furthermore, when using off-peak electricity to charge an energy storage system, energy storage is a continuous process, while energy storage efficiency fluctuates as the state of the energy storage system changes. This change in energy storage efficiency can be represented by the remaining capacity to be stored. Therefore, in the technical solution of the present invention, it is expected that the electricity demand of each electricity user is represented by the electricity power consumed at each predetermined time point.

[0169] That is, specifically, first obtain the power consumption values ​​of each electricity user at a plurality of predetermined time points, and arrange the power consumption of each electricity user at the plurality of predetermined time points into a two-dimensional input matrix according to the electricity user sample dimension and the time dimension. In this way, the data distribution information of the power consumption of each electricity user at the plurality of predetermined time points in the sample and time dimensions can be integrated. Accordingly, in a specific example, the power consumption of each electricity user at the plurality of predetermined time points is arranged into row vectors according to the time dimension to obtain a plurality of row vectors; and the plurality of row vectors are arranged into the two-dimensional input matrix according to the electricity user sample dimension. Then, the first convolutional neural network model with excellent performance in implicit correlation feature extraction is used to perform explicit spatial encoding on the two-dimensional input matrix to extract high-dimensional implicit correlation feature information of the power consumption of each electricity user at each predetermined time point, thereby obtaining an electricity consumption feature matrix.

[0170] More specifically, in step S140 and step S150, the valley power density of the power grid and the energy storage capacity of the energy storage system at the plurality of predetermined time points are obtained, and the valley power density of the power grid and the energy storage capacity of the energy storage system at the plurality of predetermined time points are respectively passed through a time series encoder comprising a one-dimensional convolutional layer to obtain a valley power density vector and an energy storage feature vector. It should be understood that when the energy storage system is charged using valley electricity, energy storage is a continuous process, while the energy storage efficiency fluctuates as the state of the energy storage system changes, and the change in energy storage efficiency can be represented by the remaining capacity to be stored. Therefore, in the technical solution of the present invention, further, the valley power density of the power grid and the energy storage capacity of the energy storage system at the plurality of predetermined time points are obtained, and they are respectively encoded in a time series encoder comprising a one-dimensional convolutional layer to mine the dynamic change characteristics of the valley power density of the power grid and the energy storage capacity of the energy storage system in the time series dimension, thereby obtaining a valley power density vector and an energy storage feature vector.

[0171] More specifically, in steps S160 and S170, the responsiveness matrix of the valley power density vector relative to the energy storage characteristic vector is calculated, and the responsiveness matrix and the power consumption characteristic matrix are fused to obtain a classification characteristic matrix. It should be understood that, considering that the valley power density characteristic of the power grid is a responsiveness characteristic of the energy storage capacity characteristic of the energy storage system, in order to strengthen the correlation characteristic information between the two to obtain a dynamic implicit change characteristic, in the technical solution of the present invention, the transfer matrix of the valley power density vector relative to the energy storage characteristic vector is further calculated as the responsiveness matrix. Here, in a specific example, the transfer matrix multiplied by the valley power density vector is equal to the energy storage characteristic vector. In this way, the responsiveness matrix and the power consumption characteristic matrix can be further fused to obtain a classification characteristic matrix. For example, the weighted sum of the responsiveness matrix and the power consumption characteristic matrix by position can be calculated to obtain the classification characteristic matrix.

[0172] More specifically, in step S180, the classification feature matrix is ​​subjected to a class condition boundary constraint synthesis based on eigenvalues ​​to obtain a corrected classification feature matrix, wherein the class condition boundary constraint synthesis based on eigenvalues ​​is performed based on the difference between the value of the natural exponential function raised to the power of the eigenvalues ​​at each position in the classification feature matrix and the inverse of the eigenvalues ​​at each position in the classification feature matrix. It should be understood that since the responsiveness matrix corresponds to the time series-time series dimension, and the power consumption feature matrix corresponds to the time series-sample dimension, when they are fused, there may be out-of-distribution eigenvalues ​​due to dimensional misalignment. Therefore, in the technical solution of the present invention, the classification feature matrix obtained by fusing the responsiveness matrix and the power consumption feature matrix is ​​subjected to a class condition boundary constraint synthesis of eigenvalues. It should be understood that the class condition boundary constraint synthesis of the eigenvalues ​​performs a boundary constraint synthesis of features by performing a rule-based structured understanding of the eigenvalues ​​and the class conditions to which they belong. This can avoid excessive fragmentation of the decision area within the classification target domain due to the out-of-distribution samples of the eigenvalue set, thereby improving the classification performance of the classification feature matrix. Furthermore, the ratio of electricity allocated to the energy storage system and that allocated to electricity users is made more reasonable.

[0173] More specifically, in step S190, the corrected classification feature matrix is ​​passed through a classifier to obtain a classification result, which is used to indicate whether the ratio of electric energy allocated to the energy storage system and the electric energy allocated to the electric user is reasonable. That is, in a specific example, the classifier processes the corrected classification feature matrix using the following formula to generate a classification result, wherein the formula is: softmax{(W n , B n):...:(W1, B1)|Project(F)}, where Project(F) represents the projection of the corrected classification feature matrix into a vector, W1 to W n is the weight matrix of each fully connected layer, B1 to B n Represents the bias matrix of each fully connected layer.

[0174] In summary, the scheduling method of the multi-energy complementary hybrid energy storage scheduling system based on the embodiment of the present invention is explained, which uses a deep neural network model of artificial intelligence technology to extract high-dimensional implicit features in the time dimension of the electricity power value of each electricity user, the valley power density of the power grid, and the energy storage capacity of the energy storage system, and uses this to perform feature fusion to judge the rationality of the proportion of electricity allocated to electricity users and energy storage systems. In addition, when fusing features, the class condition boundary constraint synthesis method of the eigenvalue is adopted to avoid excessive fragmentation of the decision area in the classification target domain due to the out-of-distribution samples of the set of eigenvalues, thereby improving the classification performance of the classification feature matrix and thus improving the accuracy of the rationality judgment.

[0175] The basic principles of the present invention have been described above in conjunction with specific embodiments. However, it should be noted that the advantages, strengths, and effects mentioned in the present invention are merely illustrative and non-limiting, and should not be construed as necessarily possessed by each embodiment of the present invention. Furthermore, the specific details disclosed above are provided for illustrative purposes and to facilitate understanding, and are not intended to be limiting. These details do not necessarily limit the present invention to being implemented using these specific details.

[0176] The block diagrams of the devices, apparatuses, equipment, and systems involved in the present invention are intended to be illustrative examples only and are not intended to require or imply that they must be connected, arranged, or configured in the manner shown in the block diagrams. As will be appreciated by those skilled in the art, these devices, apparatuses, equipment, and systems may be connected, arranged, or configured in any manner. Words such as "include," "comprise," "have," and the like are open-ended words, meaning "including but not limited to," and may be used interchangeably therewith. The words "or" and "and" used herein refer to the words "and / or" and may be used interchangeably therewith, unless the context clearly indicates otherwise. The word "such as" used herein refers to the phrase "such as but not limited to," and may be used interchangeably therewith.

[0177] It should also be noted that in the apparatus, device and method of the present invention, each component or each step can be decomposed and / or recombined, and such decomposition and / or recombination should be regarded as equivalent solutions of the present invention.

[0178] The above description of the disclosed aspects is provided to enable any person skilled in the art to make or use the present invention. Various modifications to these aspects will be readily apparent to those skilled in the art, and the general principles defined herein may be applied to other aspects without departing from the scope of the present invention. Therefore, the present invention is not intended to be limited to the aspects shown herein, but rather to be accorded the widest scope consistent with the principles and novel features disclosed herein.

[0179] The above description has been provided for the purpose of illustration and description. Furthermore, this description is not intended to limit the embodiments of the present invention to the forms disclosed herein. Although a number of example aspects and embodiments have been discussed above, those skilled in the art will recognize certain variations, modifications, alterations, additions, and sub-combinations thereof.

Claims

1. A multi-energy complementary hybrid energy storage scheduling system, characterized by: The system includes an electricity user data acquisition module, an electricity user data structuring module, an electricity consumption data encoding module, an electricity transmission and storage data acquisition module, a time series encoding module, a responsiveness estimation module, a feature fusion module, a feature distribution correction module, and a scheduling result generation module; An electricity user data collection module is used to obtain the power consumption value of each electricity user at multiple predetermined time points; An electricity user data structuring module, configured to arrange the electricity consumption of each electricity user at a plurality of predetermined time points into a two-dimensional input matrix according to the electricity user sample dimension and the time dimension; an electricity consumption data encoding module, configured to perform explicit spatial encoding on the two-dimensional input matrix using a first convolutional neural network model to obtain an electricity consumption feature matrix; The power transmission and storage data acquisition module is used to obtain the off-peak power density of the power grid and the energy storage capacity of the energy storage system at multiple predetermined time points; A time series encoding module is used to pass the valley power density of the power grid and the energy storage capacity of the energy storage system at multiple predetermined time points through a time series encoder including a one-dimensional convolutional layer to obtain a valley power density vector and an energy storage feature vector; a responsiveness estimation module, configured to calculate a responsiveness matrix of a valley power density vector relative to the energy storage characteristic vector; a feature fusion module, configured to fuse the responsiveness matrix and the power consumption feature matrix to obtain a classification feature matrix; a feature distribution correction module, configured to perform eigenvalue-based class condition boundary constraint synthesis on the classification feature matrix to obtain a corrected classification feature matrix, wherein the eigenvalue-based class condition boundary constraint synthesis is performed based on the difference between a natural exponential function value raised to the power of the eigenvalue at each position in the classification feature matrix and the inverse of the eigenvalue at each position in the classification feature matrix; The scheduling result generation module is used to pass the corrected classification feature matrix through the classifier to obtain a classification result, which is used to indicate whether the ratio of electric energy allocated to the energy storage system and allocated to electric users is reasonable.

2. The multi-energy complementary hybrid energy storage scheduling system according to claim 1 is characterized in that: The electric user data structuring module includes a vector construction unit and a matrix construction unit; a vector construction unit, configured to arrange the power consumption of each electricity user at a plurality of predetermined time points into row vectors according to the time dimension to obtain a plurality of row vectors; The matrix construction unit is used to arrange multiple row vectors into a two-dimensional input matrix according to the dimension of the electric user sample.

3. The multi-energy complementary hybrid energy storage scheduling system according to claim 2 is characterized in that: The electricity consumption data encoding module is further configured to use each layer of the first convolutional neural network model to perform the following operations on the input data in the forward pass of the layer: Perform convolution on the input data to obtain a convolution feature map; Perform mean pooling based on the local channel dimension on the convolution feature map to obtain a pooled feature map; as well as Performing nonlinear activation on the pooled feature map to obtain an activated feature map; The output of the last layer of the first convolutional neural network is a power consumption feature matrix, and the input of the first layer of the first convolutional neural network is a two-dimensional input matrix.

4. The multi-energy complementary hybrid energy storage scheduling system according to claim 3 is characterized by: The temporal coding module includes an input vector construction unit, a fully connected coding unit and a one-dimensional convolutional coding unit; An input vector construction unit, configured to arrange the valley power density of the power grid and the energy storage capacity of the energy storage system at a plurality of predetermined time points into one-dimensional input vectors according to the time dimension; The fully connected encoding unit is used to use the fully connected layer of the temporal encoder to perform fully connected encoding on the input vector using the following formula to extract the high-dimensional implicit features of the eigenvalues ​​at each position in the input vector, wherein the formula is: ,in is the input vector, is the output vector, is the weight matrix, is the bias vector, Represents matrix multiplication; The one-dimensional convolutional encoding unit is configured to perform one-dimensional convolutional encoding on the input vector using the one-dimensional convolutional layer of the temporal encoder according to the following formula to extract high-dimensional implicit correlation features between the eigenvalues ​​at each position in the input vector, wherein the formula is: ； in, a The convolution kernel is x Width in direction, F is the convolution kernel parameter vector, G is the local vector matrix operated with the convolution kernel function, w is the size of the convolution kernel.

5. The multi-energy complementary hybrid energy storage scheduling system according to claim 4 is characterized in that: The responsiveness estimation module is further configured to calculate a transfer matrix of a valley power density vector relative to an energy storage characteristic vector as the responsiveness matrix, wherein the transfer matrix multiplied by the valley power density vector equals the energy storage characteristic vector.

6. The multi-energy complementary hybrid energy storage scheduling system according to claim 5 is characterized by: The feature fusion module is further used to: fuse the responsiveness matrix and the power consumption feature matrix according to the following formula to obtain a classification feature matrix; Wherein, the formula is: ； in, M is the classification feature matrix, M 1 is the responsiveness matrix, M 2 is the electricity consumption characteristic matrix, α and β is a weighting parameter used to control the balance between the responsiveness matrix and the power consumption characteristic matrix.

7. The multi-energy complementary hybrid energy storage scheduling system according to claim 6, characterized in that: The feature distribution correction module is further used to: perform a class condition boundary constraint synthesis based on the eigenvalue on the classification feature matrix using the following formula to obtain a corrected classification feature matrix; Wherein, the formula is: ; in, Represents the eigenvalues ​​of each position in the classification feature matrix, Represents the eigenvalues ​​of each position in the corrected classification feature matrix.

8. The multi-energy complementary hybrid energy storage scheduling system according to claim 7, characterized in that: The scheduling result generation module is further configured to: the classifier processes the corrected classification feature matrix using the following formula to generate a classification result; Wherein, the formula is: ; in, Indicates that the corrected classification feature matrix is ​​projected into a vector, to is the weight matrix of each fully connected layer, to Represents the bias matrix of each fully connected layer.

9. A dispatching method for the multi-energy complementary hybrid energy storage dispatching system according to claim 1, characterized in that The steps include: S110, obtaining power consumption values ​​of each electricity user at a plurality of predetermined time points; S120, arranging the power consumption of each electricity user at a plurality of predetermined time points into a two-dimensional input matrix according to the electricity user sample dimension and the time dimension; S130, using a first convolutional neural network model to perform explicit spatial encoding on the two-dimensional input matrix to obtain a power consumption feature matrix; S140, obtaining the valley power density of the power grid and the energy storage capacity of the energy storage system at a plurality of predetermined time points; S150, passing the valley power density of the power grid and the energy storage capacity of the energy storage system at multiple predetermined time points through a temporal encoder including a one-dimensional convolutional layer to obtain a valley power density vector and an energy storage feature vector; S160, calculating the responsiveness matrix of the valley power density vector relative to the energy storage characteristic vector; S170, fusing the responsiveness matrix and the electricity usage feature matrix to obtain a classification feature matrix; S180, performing eigenvalue-based class condition boundary constraint synthesis on the classification feature matrix to obtain a corrected classification feature matrix, wherein the eigenvalue-based class condition boundary constraint synthesis is performed based on the difference between a natural exponential function value raised to a power of the eigenvalue at each position in the classification feature matrix and the inverse of the eigenvalue at each position in the classification feature matrix; S190 , passing the corrected classification feature matrix through a classifier to obtain a classification result, where the classification result is used to indicate whether the ratio of electric energy allocated to the energy storage system and to the electric energy users is reasonable.

10. The dispatching method of the multi-energy complementary hybrid energy storage dispatching system according to claim 9, characterized in that: The power consumption of each electricity user at a plurality of predetermined time points is arranged into a two-dimensional input matrix according to the electricity user sample dimension and the time dimension, including: arranging the power consumption of each electricity user at a plurality of predetermined time points into row vectors according to the time dimension to obtain a plurality of row vectors; and arranging the plurality of row vectors into a two-dimensional input matrix according to the electricity user sample dimension.

Images