Mobile energy storage auxiliary dispatching method, system and device based on multi-region power demand

Abstract

The application provides a mobile energy storage auxiliary scheduling method, system and device based on multi-region power demand, which comprises the following steps: firstly, predicting the power consumption of multiple power consumption regions in an emergency scenario; then, based on the power consumption of the multiple power consumption regions, constructing a mobile energy storage vehicle optimization scheduling model with the overall power supply process maximum benefit as the target; and finally, solving the mobile energy storage vehicle optimization scheduling model to obtain a mobile energy storage vehicle optimization scheduling scheme. The application can not only improve the satisfaction rate of the regional power demand, maintain the power supply balance of the power consumption of each region, but also improve the overall benefit of power supply.

Description

Technical Field

[0001] This invention belongs to the field of mobile energy storage dispatch, specifically relating to a mobile energy storage auxiliary dispatch method, system, and equipment based on multi-regional electricity demand. Background Technology

[0002] With the increasing demand for electricity and the further rise in electric vehicles and other electrical equipment, the power grid is facing a problem of insufficient transmission capacity. To address this, mobile energy storage technology has been proposed. Mobile energy storage systems, with their flexible spatial transfer capabilities and stable energy storage and charging characteristics, are gradually demonstrating their advantages in emergency power supply. In particular, truck-mounted mobile energy storage systems have been widely applied and promoted in post-disaster load recovery both domestically and internationally.

[0003] Existing scheduling schemes are all based on existing scenarios and cannot effectively solve the problem of untimely power supply caused by insufficient power supply for mobile energy storage vehicles. Furthermore, fixed scheduling schemes cannot maximize overall benefits. Summary of the Invention

[0004] The purpose of this invention is to address the aforementioned problems in the existing technology by providing a mobile energy storage-assisted dispatching method, system, and device based on multi-regional electricity demand.

[0005] To achieve the above objectives, the technical solution of the present invention is as follows:

[0006] In a first aspect, the present invention proposes a mobile energy storage-assisted dispatching method based on multi-regional electricity demand, comprising:

[0007] S1. In emergency scenarios, the electricity consumption of multiple power consumption areas is predicted, including commercial areas and residential areas.

[0008] S2. Based on the electricity consumption of multiple electricity consumption areas, construct an optimized scheduling model for mobile energy storage vehicles with the goal of maximizing the overall benefit of the power supply process;

[0009] S3. Solve the mobile energy storage vehicle optimal scheduling model to obtain the mobile energy storage vehicle optimal scheduling scheme.

[0010] In S2, the mobile energy storage vehicle optimization scheduling model includes:

[0011]

[0012] In the above formula, Ear represents the overall benefit of the power supply process, and Y... ind,i , Let z represent the electricity consumption of the i-th commercial area and the j-th residential area, respectively, and let N be the dispatching scheme. s Let P be the number of mobile energy storage vehicles in the s-th electricity consumption area, track be the travel route of the mobile energy storage vehicles, and P be the number of mobile energy storage vehicles in the s-th electricity consumption area.1,i P 2,j Let E and Δ be the electricity prices for the i-th commercial area and the j-th residential area, respectively; E and Δ be the electricity consumption and power loss per unit distance of the mobile energy storage vehicle, respectively; dis be the one-way distance traveled by the mobile energy storage vehicle; and P be the electricity price per unit distance of the mobile energy storage vehicle. ini The cost of electricity for charging stations, M1 and M2 are the number of commercial and residential areas respectively, and Y s Let Aviable be the electricity consumption of the s-th electricity consumption area. k Let K be the available power of the kth mobile energy storage vehicle. s This refers to the number of mobile energy storage vehicles in each electricity consumption area.

[0013] S1 includes:

[0014] S11. Using a self-attention mechanism, the historical electricity consumption sequences of each electricity consumption area are exchanged as follows:

[0015]

[0016] In the above formula, This represents the feature mapping result of the historical electricity consumption sequence for the i-th electricity consumption area. Embedding is an independent feature mapping operation. Let L be the l-th data set in the historical electricity consumption sequence of the i-th electricity consumption area, where L is the length of the historical electricity consumption sequence.

[0017] S12. Use layer normalization to unify the feature channel distribution of different variables:

[0018]

[0019] In the above formula, LayerNorm is the layer normalization operation, and H i After layer normalization operation Mean is the average value, and var is the variance.

[0020] S13. Perform fully connected feature encoding in the feedforward network, and predict the electricity consumption of each power consumption area through projection layer mapping:

[0021] Y i {L:L+m} =Projection(H i )

[0022] In the above formula, Y i {L:L+m} This represents the predicted electricity consumption result for the i-th electricity consumption area with a sequence length of m. Projection is a mapping operation.

[0023] The S3 method uses the DDQN algorithm to solve the mobile energy storage vehicle optimal scheduling model, including:

[0024] S31. Transform the mobile energy storage vehicle optimization scheduling model into an energy storage vehicle quantity optimization model and a path planning model;

[0025] S32. Construct the equivalent Markov state transition process for the path planning model:

[0026] max Qπ(s,a;θ)=E[R1+γR2+…|S,A,π;θ]

[0027] In the above formula, Q π (s,a;θ) represents the cumulative expected reward under path planning, where s, a, and θ are the state, action, and network parameters, respectively, R is the reward for each corresponding action, S and A are the state space and action space, respectively, the action space represents the choice at each intersection, π is the established path planning strategy, γ is the decay coefficient, and E[·] is the expectation of x;

[0028] S33. The DDQN algorithm is used for iterative solution to obtain the path planning strategy π, which is the track in the objective function.

[0029] Secondly, this invention proposes a mobile energy storage-assisted scheduling system based on multi-regional electricity demand, including an electricity consumption prediction module, a scheduling model construction module, and a scheduling model solving module;

[0030] The electricity consumption prediction module is used to predict the electricity consumption of multiple electricity consumption areas in emergency scenarios, including commercial areas and residential areas.

[0031] The scheduling model construction module is used to construct an optimized scheduling model for mobile energy storage vehicles based on the electricity consumption of multiple electricity consumption areas, with the goal of maximizing the overall benefit of the power supply process.

[0032] The scheduling model solving module is used to solve the mobile energy storage vehicle optimal scheduling model to obtain the mobile energy storage vehicle optimal scheduling scheme.

[0033] The mobile energy storage vehicle optimization scheduling model includes:

[0034]

[0035] In the above formula, Ear represents the overall benefit of the power supply process, and Y... ind,i , Let z represent the electricity consumption of the i-th commercial area and the j-th residential area, respectively, and let N be the dispatching scheme. s Let P be the number of mobile energy storage vehicles in the s-th electricity consumption area, track be the travel route of the mobile energy storage vehicles, and P be the number of mobile energy storage vehicles in the s-th electricity consumption area. 1,i P 2,jLet E and Δ be the electricity prices for the i-th commercial area and the j-th residential area, respectively; E and Δ be the electricity consumption and power loss per unit distance of the mobile energy storage vehicle, respectively; dis be the one-way distance traveled by the mobile energy storage vehicle; and P be the electricity price per unit distance of the mobile energy storage vehicle. ini The cost of electricity for charging stations, M1 and M2 are the number of commercial and residential areas respectively, and Y s Let Aviable be the electricity consumption of the s-th electricity consumption area. k Let K be the available power of the kth mobile energy storage vehicle. s This refers to the number of mobile energy storage vehicles in each electricity consumption area.

[0036] The electricity consumption prediction module uses the following steps to predict the electricity consumption of multiple electricity consumption areas:

[0037] A. Using a self-attention mechanism, the historical electricity consumption sequences of each electricity consumption area are exchanged as follows:

[0038]

[0039] In the above formula, This represents the feature mapping result of the historical electricity consumption sequence for the i-th electricity consumption area. Embedding is an independent feature mapping operation. Let L be the l-th data set in the historical electricity consumption sequence of the i-th electricity consumption area, where L is the length of the historical electricity consumption sequence.

[0040] B. Use layer normalization to unify the feature channel distribution of different variables:

[0041]

[0042] In the above formula, LayerNorm is the layer normalization operation, and H i After layer normalization operation Mean is the average value, and var is the variance.

[0043] C. Perform fully connected feature encoding in the feedforward network, and predict the electricity consumption of each power consumption area through projection layer mapping:

[0044] Y i {L:L+m} =Projection(H i )

[0045] In the above formula, Y i {L:L+m} This represents the predicted electricity consumption result for the i-th electricity consumption area with a sequence length of m. Projection is a mapping operation.

[0046] The scheduling model solving module uses the DDQN algorithm to solve the mobile energy storage vehicle optimal scheduling model, including:

[0047] a. Transform the mobile energy storage vehicle optimization scheduling model into an energy storage vehicle quantity optimization model and a path planning model;

[0048] b. Construct the equivalent Markov state transition process for the path planning model:

[0049] max Qπ(s,a;θ)=E[R1+γR2+…|S,A,π;θ]

[0050] In the above formula, Q π (s,a;θ) represents the cumulative expected reward under path planning, where s, a, and θ are the state, action, and network parameters, respectively, R is the reward for each corresponding action, S and A are the state space and action space, respectively, the action space represents the choice at each intersection, π is the established path planning strategy, γ is the decay coefficient, and E[·] is the expectation of x;

[0051] c. The DDQN algorithm is used for iterative solution to obtain the path planning strategy π, which is the track in the objective function.

[0052] Thirdly, this invention proposes a mobile energy storage auxiliary dispatching device based on multi-regional electricity demand, including a memory and a processor;

[0053] The memory is used to store computer program code and transmit the computer program code to the processor;

[0054] The processor is configured to execute the aforementioned method according to instructions in the computer program code.

[0055] Fourthly, the present invention provides a computer-readable storage medium on which a computer program is stored, and which, when executed by a processor, implements the aforementioned method.

[0056] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0057] 1. This invention discloses a mobile energy storage-assisted scheduling method based on multi-regional electricity demand. First, it predicts the electricity consumption of multiple electricity-consuming regions in an emergency scenario. Then, based on the electricity consumption of these regions, it constructs an optimized scheduling model for mobile energy storage vehicles, aiming to maximize the overall benefit of the power supply process. Finally, it solves the optimized scheduling model to obtain an optimized scheduling scheme for mobile energy storage vehicles. This method starts from the perspective of actual regional electricity demand, optimizing the scheduling scheme of mobile energy storage vehicles while ensuring the electricity demand of each region. This not only improves the satisfaction rate of regional electricity demand and maintains the power supply balance in each region, but also enhances the overall benefit of power supply.

[0058] 2. This invention provides a mobile energy storage-assisted scheduling method based on multi-regional electricity demand. For predicting electricity consumption in each region, it employs a self-attention mechanism to facilitate information exchange between historical electricity consumption data of the four regions. Layer normalization is used to unify the feature distribution of different variables, and fully connected feature encoding is performed in a feedforward network. The prediction is then directly performed through a projection layer. This method effectively improves the accuracy of regional electricity consumption prediction. Attached Figure Description

[0059] Figure 1 This is a diagram of the regional network model used in Example 1.

[0060] Figure 2 This is a flowchart of the method described in Example 1.

[0061] Figure 3 This is a comparison diagram of the overall benefits of the power supply process between the method described in this invention and the fixed-route method.

[0062] Figure 4 This is a structural block diagram of the system described in Example 2.

[0063] Figure 5 This is a structural block diagram of the device described in Example 3. Detailed Implementation

[0064] The present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0065] Example 1:

[0066] This embodiment is for Figure 1 The regional network model diagram shown (containing four power consumption areas, namely industrial park, residential area 1, residential area 2, and residential area 3) implements a mobile energy storage-assisted dispatch method based on multi-area power demand, such as... Figure 2 As shown, the specific steps are as follows:

[0067] 1. In a power outage scenario, obtain historical electricity consumption data D for four electricity consumption areas recorded in the power grid system. ind ={d ind,1 ,…,d ind,L}、

[0068] 2. Using a self-attention mechanism, the historical electricity consumption sequences of each electricity consumption area are exchanged as follows:

[0069]

[0070] In the above formula, These are the feature mapping results for historical electricity consumption sequences of the industrial park, residential area 1, residential area 2, and residential area 3, respectively. Embedding is an independent feature mapping operation. These are the l-th data sets in the historical electricity consumption sequences of the industrial park, residential area 1, residential area 2, and residential area 3, respectively, where L is the length of the historical electricity consumption sequence.

[0071] 3. Use layer normalization to unify the feature channel distribution of different variables:

[0072]

[0073] In the above formula, LayerNorm is the layer normalization operation, and H i After layer normalization operation Mean is the average value, and var is the variance.

[0074] 4. Perform fully connected feature encoding in the feedforward network, and predict the electricity consumption of each power consumption area through projection layer mapping:

[0075] Y i {L:L+m} =Projection(H i )

[0076] In the above formula, Y i {L:L+m} This represents the predicted electricity consumption result for the i-th electricity consumption area with a sequence length of m. Projection is a mapping operation.

[0077] 5. Based on the regional network model, derive the overall revenue model of the power supply process.

[0078] The electricity price in the commercial area is P1, and the electricity price in the residential area is P2. The mobile energy storage vehicle is a pure electric vehicle, and its electricity consumption per kilometer is E. However, energy storage losses during the movement are negligible. The speed of the mobile energy storage vehicle on the main line is v1, and its speed on the branch lines is v2. The expenditure for supplying electricity to these four areas using the mobile vehicle includes two parts: the electricity used by the electric vehicle during its movement and the losses of the energy storage vehicle, as shown below:

[0079] consume = 2·(E+Δ)·dis

[0080] In the above formula, consume represents the total expenditure, Δ represents the loss per kilometer of the energy storage vehicle, and dis represents the one-way distance traveled by the mobile energy storage vehicle.

[0081] The revenue breakdown is the electricity fees collected in these four areas, expressed as follows:

[0082]

[0083] In the above formula, Y ind , These represent the electricity consumption of the commercial area and the i-th residential area, respectively.

[0084] Therefore, the overall benefit of the mobile energy storage vehicle in the process of providing power is expressed as follows:

[0085]

[0086] In the above formula, P ini The cost of electricity for charging stations.

[0087] 6. To ensure the electricity demand of each region, construct the following optimized scheduling model for mobile energy storage vehicles:

[0088]

[0089] In the above formula, z represents the scheduling scheme, and N i Let represent the number of mobile energy storage vehicles in the i-th power consumption area, and track represent the travel route of the mobile energy storage vehicles.

[0090] 7. The above scheduling model is transformed into an energy storage vehicle quantity optimization model and a route planning model, which are as follows:

[0091]

[0092] Since the number of energy storage vehicles N only meets the electricity demand of the industrial park, residential area 1, residential area 2, and residential area 3, this part can be transformed into an equivalent constraint. Therefore, the optimal solution of the optimization model (P1) can be directly obtained, i.e., N. s =Y s / Aviable, where Aviable is the available power of each mobile energy storage vehicle. The energy storage vehicle travel routes in the optimization model (P2) still require further analysis; therefore, the optimization model (OP) can be simplified to:

[0093]

[0094] 8. Construct the equivalent Markov state transition process for the path planning model (P3):

[0095] max Q π (s,a;θ)=E[R1+γR2+…|S,A,π;θ]

[0096] In the above formula, Q π(s,a;θ) represents the cumulative expected reward under path planning, where s, a, and θ are the state, action, and network parameters, respectively, R is the reward for each corresponding action, S and A are the state space and action space, respectively, the action space represents the choice at each intersection, including four actions: up, down, left, and right, π is the defined path planning strategy, γ is the decay coefficient, and E[·] is the expectation of x.

[0097] 9. The DDQN algorithm is used for iterative solution to obtain the path planning strategy π, i.e., track in the objective function. The network selection and update methods are as follows:

[0098] Y t Q =R t+1 +γQ(S t+1 ,argmaxQ(S t+1 ,a;θ t );θ t )

[0099] Y t DoubleQ =R t+1 +γQ(S t+1 ,argmaxQ(S t+1 ,a;θ t );θ t ′)

[0100] In the above formula, Y t Q Y t DoubleQ Let R be the target network at time t, and the optimization objective of the estimated network. t+1 S is the reward value obtained after performing the action at time t+1. t+1 Let Q(S) be the state at time t+1. t+1 ,a;θ t ) represents the Q-value of the target network, and θ represents the Q-value of the target network. t θ t ' and ' are the parameters of the target network and the estimated network at time t, respectively.

[0101] To verify the effectiveness of the method described in this invention, the method described in Example 1 was compared with the fixed-route method. The overall benefits of the power supply process for both methods were calculated, and the results are as follows: Figure 3 As shown, the overall benefits of the method described in this invention are significantly higher than those of the stationary driving method.

[0102] Example 2:

[0103] A mobile energy storage-assisted dispatch system based on multi-regional electricity demand, such as Figure 4As shown, it includes a power consumption prediction module, a scheduling model construction module, and a scheduling model solving module.

[0104] The electricity consumption prediction module is used to predict the electricity consumption of multiple electricity consumption areas in emergency scenarios by employing the following steps:

[0105] A. Using a self-attention mechanism, the historical electricity consumption sequences of each electricity consumption area are exchanged as follows:

[0106]

[0107] In the above formula, This represents the feature mapping result of the historical electricity consumption sequence for the i-th electricity consumption area. Embedding is an independent feature mapping operation, and D... hi,l Let L be the l-th data set in the historical electricity consumption sequence of the i-th electricity consumption area, where L is the length of the historical electricity consumption sequence.

[0108] B. Use layer normalization to unify the feature channel distribution of different variables:

[0109]

[0110] In the above formula, LayerNorm is the layer normalization operation, and H i After layer normalization operation Mean is the average value, and var is the variance.

[0111] C. Perform fully connected feature encoding in the feedforward network, and predict the electricity consumption of each power consumption area through projection layer mapping:

[0112] Y i {L:L+m} =Projection(H i )

[0113] In the above formula, Y i {L:L+m} This represents the predicted electricity consumption result for the i-th electricity consumption area with a sequence length of m. Projection is a mapping operation.

[0114] The scheduling model construction module is used to construct an optimized scheduling model for mobile energy storage vehicles based on the electricity consumption of multiple electricity consumption areas, with the goal of maximizing the overall benefit of the power supply process:

[0115]

[0116] In the above formula, Ear represents the overall benefit of the power supply process, and Y... ind,i , Let z represent the electricity consumption of the i-th commercial area and the j-th residential area, respectively, and let N be the dispatching scheme. sLet P be the number of mobile energy storage vehicles in the s-th electricity consumption area, track be the travel route of the mobile energy storage vehicles, and P be the number of mobile energy storage vehicles in the s-th electricity consumption area. 1,i P 2,j Let E and Δ be the electricity prices for the i-th commercial area and the j-th residential area, respectively; E and Δ be the electricity consumption and power loss per unit distance of the mobile energy storage vehicle, respectively; dis be the one-way distance traveled by the mobile energy storage vehicle; and P be the electricity price per unit distance of the mobile energy storage vehicle. ini The cost of electricity for charging stations, M1 and M2 are the number of commercial and residential areas respectively, and Y s Let Aviable be the electricity consumption of the s-th electricity consumption area. k Let K be the available power of the kth mobile energy storage vehicle. s This refers to the number of mobile energy storage vehicles in each electricity consumption area.

[0117] The scheduling model solving module uses the DDQN algorithm to solve the mobile energy storage vehicle optimal scheduling model, and obtains the mobile energy storage vehicle optimal scheduling scheme. The solution process includes:

[0118] a. Transform the mobile energy storage vehicle optimization scheduling model into an energy storage vehicle quantity optimization model and a path planning model;

[0119] b. Construct the equivalent Markov state transition process for the path planning model:

[0120] max Qπ(s,a;θ)=E[R1+γR2+…|S,A,π;θ]

[0121] In the above formula, Q π (s,a;θ) represents the cumulative expected reward under path planning, where s, a, and θ are the state, action, and network parameters, respectively, R is the reward for each corresponding action, S and A are the state space and action space, respectively, the action space represents the choice at each intersection, π is the defined path planning strategy, γ is the decay coefficient, and E[x] is the expectation of x.

[0122] c. The DDQN algorithm is used for iterative solution to obtain the path planning strategy π, which is the track in the objective function.

[0123] Example 3:

[0124] A mobile energy storage auxiliary dispatching device based on multi-regional electricity demand, such as Figure 5 As shown, it includes memory and processor;

[0125] The memory is used to store computer program code and transmit the computer program code to the processor;

[0126] The processor is configured to execute the method described in Embodiment 1 according to instructions in the computer program code.

[0127] Example 4:

[0128] A computer-readable storage medium storing a computer program that, when executed by a processor, implements the method described in Embodiment 1.

Claims

1. A mobile energy storage-assisted dispatching method based on multi-regional electricity demand, characterized in that, The method includes: S1. In emergency scenarios, the electricity consumption of multiple power consumption areas is predicted, including commercial areas and residential areas. S2. Based on the electricity consumption of multiple electricity consumption areas, construct an optimal scheduling model for mobile energy storage vehicles with the objective of maximizing the overall benefit of the power supply process. The optimal scheduling model for mobile energy storage vehicles includes: ； ； In the above formula, For the overall benefit of the power supply process, , These represent the electricity consumption of the i-th commercial area and the j-th residential area, respectively. For the scheduling scheme, Let be the number of mobile energy storage vehicles in the s-th electricity consumption area. The route for the mobile energy storage vehicle. , Let be the electricity prices for the i-th commercial area and the j-th residential area, respectively. , These represent the electricity consumption and power loss per unit distance for mobile energy storage vehicles. This refers to the one-way distance traveled by the mobile energy storage vehicle. The cost of electricity for charging at charging stations , The number of commercial areas and residential areas, respectively. Let be the electricity consumption of the s-th electricity consumption area. Let k be the available power for the mobile energy storage vehicle. The number of mobile energy storage vehicles in each electricity consumption area; S3. Solve the mobile energy storage vehicle optimal scheduling model to obtain the mobile energy storage vehicle optimal scheduling scheme.

2. The mobile energy storage-assisted dispatching method based on multi-regional electricity demand according to claim 1, characterized in that, S1 includes: S11. Using a self-attention mechanism, the historical electricity consumption sequences of each electricity consumption area are exchanged as follows: ； In the above formula, This represents the feature mapping result of the historical electricity consumption sequence for the i-th electricity consumption area. For independent feature mapping operations, This is the l-th set of data in the historical electricity consumption sequence of the i-th electricity consumption area. The length of the historical electricity consumption sequence; S12. Use layer normalization to unify the feature channel distribution of different variables: ； In the above formula, For layer normalization operation, After layer normalization operation , To take the average value, To obtain the variance; S13. Perform fully connected feature encoding in the feedforward network, and predict the electricity consumption of each power consumption area through projection layer mapping: ； In the above formula, For a sequence length of The predicted electricity consumption results for the i-th electricity consumption area This is a mapping operation.

3. The mobile energy storage-assisted dispatching method based on multi-regional electricity demand according to claim 1, characterized in that, The S3 method uses the DDQN algorithm to solve the mobile energy storage vehicle optimal scheduling model, including: S31. Transform the mobile energy storage vehicle optimization scheduling model into an energy storage vehicle quantity optimization model and a path planning model; S32. Construct the equivalent Markov state transition process for the path planning model: ； In the above formula, The cumulative expected reward under path planning. , , These are state, action, and network parameters, respectively. The rewards for each corresponding action, , These are the state space and the action space, respectively. The action space represents the choices made at each intersection. The established path planning strategy The attenuation coefficient is... To Seeking expectations; S33. The DDQN algorithm is used for iterative solution to obtain the path planning strategy. That is, in the objective function .

4. A mobile energy storage-assisted dispatching system based on multi-regional electricity demand, characterized in that, The system includes a power consumption prediction module, a scheduling model construction module, and a scheduling model solving module; The electricity consumption prediction module is used to predict the electricity consumption of multiple electricity consumption areas in emergency scenarios, including commercial areas and residential areas. The scheduling model construction module is used to construct an optimized scheduling model for mobile energy storage vehicles based on the electricity consumption of multiple electricity consumption areas, with the goal of maximizing the overall benefit of the power supply process. The optimized scheduling model for mobile energy storage vehicles includes: ； ； In the above formula, For the overall benefit of the power supply process, , These represent the electricity consumption of the i-th commercial area and the j-th residential area, respectively. For the scheduling scheme, Let be the number of mobile energy storage vehicles in the s-th electricity consumption area. The route for the mobile energy storage vehicle. , Let be the electricity prices for the i-th commercial area and the j-th residential area, respectively. , These represent the electricity consumption and power loss per unit distance for mobile energy storage vehicles. This refers to the one-way distance traveled by the mobile energy storage vehicle. The cost of electricity for charging at charging stations , The number of commercial areas and residential areas, respectively. Let be the electricity consumption of the s-th electricity consumption area. Let k be the available power for the mobile energy storage vehicle. The number of mobile energy storage vehicles in each electricity consumption area; The scheduling model solving module is used to solve the mobile energy storage vehicle optimal scheduling model to obtain the mobile energy storage vehicle optimal scheduling scheme.

5. A mobile energy storage-assisted dispatching system based on multi-regional electricity demand according to claim 4, characterized in that, The electricity consumption prediction module uses the following steps to predict the electricity consumption of multiple electricity consumption areas: A. Using a self-attention mechanism, the historical electricity consumption sequences of each electricity consumption area are exchanged as follows: ； In the above formula, This represents the feature mapping result of the historical electricity consumption sequence for the i-th electricity consumption area. For independent feature mapping operations, This is the l-th set of data in the historical electricity consumption sequence of the i-th electricity consumption area. The length of the historical electricity consumption sequence; B. Use layer normalization to unify the feature channel distribution of different variables: ； In the above formula, For layer normalization operation, After layer normalization operation , To take the average value, To obtain the variance; C. Perform fully connected feature encoding in the feedforward network, and predict the electricity consumption of each power consumption area through projection layer mapping: ； In the above formula, For a sequence length of The predicted electricity consumption results for the i-th electricity consumption area This is a mapping operation.

6. A mobile energy storage-assisted dispatching system based on multi-regional electricity demand according to claim 4, characterized in that, The scheduling model solving module uses the DDQN algorithm to solve the mobile energy storage vehicle optimal scheduling model. The solution process includes: a. Transform the mobile energy storage vehicle optimization scheduling model into an energy storage vehicle quantity optimization model and a path planning model; b. Construct the equivalent Markov state transition process for the path planning model: ； In the above formula, The cumulative expected reward under path planning. , , These are state, action, and network parameters, respectively. The rewards for each corresponding action, , These are the state space and the action space, respectively. The action space represents the choices made at each intersection. The established path planning strategy The attenuation coefficient is... To Seeking expectations; c. The DDQN algorithm is used for iterative solution to obtain the path planning strategy. That is, in the objective function .

7. A mobile energy storage auxiliary dispatching device based on multi-regional electricity demand, characterized in that, The device includes a memory and a processor; The memory is used to store computer program code and transmit the computer program code to the processor; The processor is configured to execute the method as described in any one of claims 1-3 according to instructions in the computer program code.

8. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program that, when executed by a processor, implements the method as described in any one of claims 1-3.

Images