Cooperative scheduling method and system for novel power distribution network

Abstract

The invention discloses a cooperative scheduling method and system for a novel power distribution network, and relates to the technical field of electric power overall planning, and the method comprises the steps: collecting historical source network load storage data of the novel power distribution network; according to the historical energy data and the historical weather data, calculating to obtain an energy data predicted value and a load data predicted value of the novel power distribution network; carrying out fusion processing on the historical energy storage data, and calculating to obtain an energy storage data predicted value of the novel power distribution network; constructing a novel power distribution network source network load storage integrated operation model; and solving the operation model to obtain a power distribution strategy, and scheduling the novel power distribution network according to the power distribution strategy. By adopting the embodiment of the invention, the distributed energy storage power, the distributed energy power and the load power can be dispatched cooperatively, the potential of the distributed energy and the distributed energy storage in the novel power distribution network can be brought into full play, and the source-network-load-storage integrated operation of the novel power distribution network can be realized.

Description

Technical Field

[0001] The present invention relates to the technical field of power coordination, and in particular to a new coordinated dispatching method and system for a power distribution network. Background Art

[0002] With the increasing number of distributed energy sources, energy storage systems and load types in distribution networks, the above-mentioned resources have different characteristics, response speeds and uncertainties. By coordinating the source, grid, load and storage, the integrated operation of the source, grid, load and storage of the new distribution network can be realized, which is of great significance to improving the operation quality of the distribution network and supporting the operation of the power grid.

[0003] Traditional dispatching methods are unable to effectively handle these diverse resources, and the prediction of resources is not accurate enough, resulting in a significant increase in the overall dispatching complexity. The generated dispatching plans are difficult to meet the requirements in actual operation, and may even lead to instability in the distribution network. Summary of the invention

[0004] The present invention provides a coordinated dispatching method and system for a new distribution network, which can coordinate the dispatching of distributed energy storage power, distributed energy power and load power, give full play to the potential of distributed energy and distributed energy storage in the new distribution network, and realize the integrated operation of source, grid, load and storage of the new distribution network.

[0005] In order to solve the technical problem that the prior art cannot effectively coordinate the dispatch of various resources in a new distribution network, an embodiment of the present invention provides a new distribution network coordinated dispatch method, including:

[0006] Collecting historical source, grid, load and storage data of the new distribution network; the historical source, grid, load and storage data includes historical energy data, historical distribution network price data, historical weather data and historical energy storage data;

[0007] Calculate the energy data forecast value and the load data forecast value of the new distribution network according to the historical energy data and the historical weather data;

[0008] Performing fusion processing on the historical energy storage data to calculate and obtain a predicted value of energy storage data of the new distribution network;

[0009] According to the historical distribution network price data, the energy data forecast value, the load data forecast value and the energy storage data forecast value, a new distribution network source-grid-load-storage integrated operation model is constructed;

[0010] The operation model is solved to obtain a power allocation strategy, and the new distribution network is dispatched according to the power allocation strategy; the power allocation strategy includes distributed energy storage power, distributed photovoltaic power and load power.

[0011] As an improvement of the above solution, the energy data forecast value and the load data forecast value of the new distribution network are calculated based on the historical energy data and the historical weather data, including:

[0012] Obtaining energy and weather data sequences according to the historical energy data and the historical weather data;

[0013] A multi-head self-attention mechanism is used to calculate the dependency between the energy and weather data sequences to obtain a multi-head self-attention score;

[0014] Establishing an energy load forecasting model based on the multi-head self-attention scores through a feedforward neural network;

[0015] Solve the energy load forecasting model and calculate the energy data forecast value and load data forecast value of the new distribution network.

[0016] As an improvement of the above scheme, the multi-head self-attention mechanism is used to calculate the dependency between the energy and weather data sequences to obtain the multi-head self-attention score, including:

[0017] adding position codes to the energy and weather data sequences according to time information of the energy and weather data sequences;

[0018] A multi-head self-attention mechanism is adopted to calculate the dependency of the energy and weather data sequences with the added position encoding, and obtain a multi-head self-attention score.

[0019] As an improvement of the above solution, the historical energy storage data is fused and processed to calculate the predicted value of energy storage data of the new distribution network, including:

[0020] Performing fusion processing on the historical energy storage data to obtain fused energy storage data;

[0021] Establishing an energy storage prediction model based on the fused energy storage data;

[0022] Solve the energy storage prediction model and calculate the energy storage data prediction value of the new distribution network.

[0023] As an improvement of the above solution, the energy storage prediction model is established according to the fused energy storage data, including:

[0024] Adopting a multimodal AI-based algorithm, and establishing an energy storage charging power prediction model and an energy storage discharging power prediction model respectively according to the fused energy storage data;

[0025] An energy storage prediction model is obtained according to the energy storage charging power prediction model and the energy storage discharging power prediction model.

[0026] As an improvement of the above solution, the multimodal AI algorithm is used to establish an energy storage charging power prediction model and an energy storage discharging power prediction model according to the fused energy storage data, including:

[0027] pass Establish a storage charging power prediction model; where Z C (t) is the fusion energy storage data at time t, f C (Z C (t)) is the energy storage charging power prediction model built based on the multimodal AI algorithm. is the distributed energy storage charging power calculated based on the fused energy storage data at time t;

[0028] pass Establish a storage charging power prediction model; where Z C (t) is the fusion energy storage data at time t, f F (Z C (t)) is the energy storage discharge power prediction model built based on the multimodal AI algorithm. is the distributed energy storage discharge power calculated based on the fused energy storage data at time t.

[0029] As an improvement of the above scheme, the operation model of a new distribution network integrating source, grid, load and storage is constructed according to the historical distribution network price data, the energy data prediction value, the load data prediction value and the energy storage data prediction value, including:

[0030] Obtaining the total operating cost of the new distribution network according to the historical distribution network price data, the energy data prediction value and the energy storage data prediction value;

[0031] Obtaining a comprehensive power grid response deviation of the new distribution network according to the energy data prediction value, the load data prediction value and the energy storage data prediction value;

[0032] generating an objective function according to the total operating cost and the comprehensive grid response deviation;

[0033] Establishing constraints of the objective function;

[0034] According to the objective function and the constraints, a new distribution network source-grid-load-storage integrated operation model is constructed.

[0035] As an improvement of the above solution, the constraint conditions for establishing the objective function include:

[0036] According to the upper and lower limit constraints of the distributed energy storage charging power, a first constraint condition is established;

[0037] According to the upper and lower limit constraints of the distributed energy storage discharge power, a second constraint condition is established;

[0038] According to the upper and lower limits of distributed energy photovoltaic power, the third constraint condition is established;

[0039] According to the upper and lower limit constraints of distributed energy load power, the fourth constraint condition is established;

[0040] The constraint conditions of the objective function are established according to the first constraint condition, the second constraint condition, the third constraint condition and the fourth constraint condition.

[0041] As an improvement of the above solution, solving the operation model to obtain a power allocation strategy, and scheduling the new distribution network according to the power allocation strategy includes:

[0042] Establishing a loss function of the running model;

[0043] The operation model is solved with the goal of minimizing the loss function to obtain a power allocation strategy; in the process of solving the operation model, the learnable parameters in the solution model are updated at each time step;

[0044] The new power distribution network is dispatched according to the power allocation strategy.

[0045] As an improvement of the above solution, the operation model is solved with the goal of minimizing the loss function to obtain a power allocation strategy; in the process of solving the operation model, the learnable parameters in the operation model are updated at each time step, including:

[0046] According to the loss function, a reward function is obtained;

[0047] Calculating the expectation of the reward function and the learning rate of the running model, and optimizing the learnable parameters in the running model at each time step;

[0048] Based on the continuously optimized learnable parameters and with the goal of minimizing the loss function, the operation model is solved to obtain a power allocation strategy.

[0049] The embodiment of the present invention further provides a novel coordinated dispatching system for distribution networks, including:

[0050] A data acquisition module is used to collect historical source-grid-load-storage data of the new distribution network; the historical source-grid-load-storage data includes historical energy data, historical distribution network price data, historical weather data and historical energy storage data;

[0051] An energy data prediction module, used to calculate the energy data prediction value and the load data prediction value of the new distribution network according to the historical energy data and the historical weather data;

[0052] An energy storage data prediction module is used to perform fusion processing on the historical energy storage data and calculate the energy storage data prediction value of the new distribution network;

[0053] A model building module, used to build a new distribution network source-grid-load-storage integrated operation model according to the historical distribution network price data, the energy data forecast value, the load data forecast value and the energy storage data forecast value;

[0054] A strategy generation module is used to solve the operation model, obtain a power allocation strategy, and dispatch the new distribution network according to the power allocation strategy; the power allocation strategy includes distributed energy storage power, distributed energy power and load power.

[0055] Compared with the prior art, the present invention discloses a new type of distribution network collaborative dispatching method and system, which collects historical source-grid-load-storage data of the new type of distribution network; the historical source-grid-load-storage data includes historical energy data, historical distribution network price data, historical weather data and historical energy storage data; according to the historical energy data and the historical weather data, the energy data prediction value and the load data prediction value of the new type of distribution network are calculated; the historical energy storage data are fused and processed to calculate the energy storage data prediction value of the new type of distribution network; according to the historical distribution network price data, the energy data prediction value, the load data prediction value and the energy storage data prediction value, an operation model of the new type of distribution network source-grid-load-storage integration is constructed; the operation model is solved to obtain a power allocation strategy, and the new type of distribution network is dispatched according to the power allocation strategy; the power allocation strategy includes distributed energy storage power, distributed photovoltaic power and load power. By adopting the embodiment of the present invention, the distributed energy storage power, distributed energy power and load power can be coordinated and dispatched, the potential of distributed energy and distributed energy storage in the new type of distribution network can be fully utilized, and the source-grid-load-storage integrated operation of the new type of distribution network can be realized. BRIEF DESCRIPTION OF THE DRAWINGS

[0056] Figure 1 It is a schematic diagram of the steps of a new coordinated dispatching method for distribution networks provided by an embodiment of the present invention;

[0057] Figure 2 It is a structural diagram of a new type of coordinated dispatching system for distribution network provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0058] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0059] In the description of the specification and claims, it is to be understood that the terms first, second, etc. in the specification and claims are only used for the purpose of describing the same technical features, and cannot be understood as indicating or implying relative importance or implicitly indicating the number of technical features indicated, nor necessarily describing the order or time sequence. The terms are interchangeable where appropriate. Therefore, a feature defined as "first" or "second" may explicitly or implicitly include at least one of the features.

[0060] The embodiment of the present invention provides a new coordinated dispatching method for distribution network, see Figure 1 In this embodiment, the novel coordinated dispatching method of the distribution network is specifically performed through steps S1 to S5:

[0061] S1. Collect historical source, grid, load and storage data of the new distribution network; the historical source, grid, load and storage data include historical energy data, historical distribution network price data, historical weather data and historical energy storage data.

[0062] Exemplarily, the historical energy data includes historical active power data and historical load data of distributed energy; the historical distribution network price data includes historical electricity prices, historical photovoltaic grid-connected prices and energy storage discharge prices; the historical energy storage data includes historical SOC data, historical charging power data and historical discharge power data of distributed energy storage.

[0063] It should be noted that the historical source network load and storage data includes data from several historical moments, and the data includes time information, which can reflect the operation status of the new distribution network at different times.

[0064] S2. Calculate the energy data forecast value and the load data forecast value of the new distribution network based on the historical energy data and the historical weather data.

[0065] It should be noted that weather will have an impact on energy consumption in the new distribution network. The energy data forecast value and load data forecast value of the new distribution network generated based on the historical energy data and the historical weather data are more accurate and explainable than only extracting the features of historical energy data on a time scale to predict future energy data.

[0066] S3. Perform fusion processing on the historical energy storage data to calculate and obtain a predicted value of energy storage data of the new distribution network.

[0067] In an embodiment of the present invention, by fusing the historical energy storage data, the correlation between the historical energy storage data can be deeply mined. For example, by mining the correlation between SOC and the charging and discharging curves, the functional relationship between the distributed energy storage charging and discharging curves and SOC is given, and the distributed energy storage charging and discharging active power curves are accurately predicted. Compared with the technical solution in the prior art that assumes that the charging power and discharging power of the distributed energy storage are fixed values ​​and predicts the energy storage data, the predicted value calculated by the solution provided in the embodiment of the present invention is more accurate.

[0068] S4. Construct a new distribution network source-grid-load-storage integrated operation model based on the historical distribution network price data, the energy data forecast value, the load data forecast value and the energy storage data forecast value.

[0069] In an embodiment of the present invention, the operating model is used to calculate how the power of distributed energy and distributed energy storage in the new distribution network should be scheduled and allocated based on the historical distribution network price data, the energy data prediction value, the load data prediction value and the energy storage data prediction value, so as to achieve requirements for the economy or stability of the distribution network.

[0070] S5. Solve the operation model to obtain a power allocation strategy, and dispatch the new distribution network according to the power allocation strategy; the power allocation strategy includes distributed energy storage power, distributed photovoltaic power and load power.

[0071] In an embodiment of the present invention, preferably, a GPT model based on artificial intelligence is used to solve the operation model. By introducing the GPT model, decision-making and scheduling can be made more intelligent and more adaptable.

[0072] As a preferred implementation, step S2, calculating the energy data forecast value and the load data forecast value of the new distribution network according to the historical energy data and the historical weather data, includes:

[0073] Obtaining energy and weather data sequences according to the historical energy data and the historical weather data;

[0074] A multi-head self-attention mechanism is used to calculate the dependency between the energy and weather data sequences to obtain a multi-head self-attention score;

[0075] Establishing an energy load forecasting model based on the multi-head self-attention scores through a feedforward neural network;

[0076] Solve the energy load forecasting model and calculate the energy data forecast value and load data forecast value of the new distribution network.

[0077] Preferably, in an embodiment of the present invention, the energy data prediction value of the new distribution network is calculated based on the Transformer architecture, and a multi-head self-attention mechanism is used in the calculation process to capture the dependency between the features of the historical energy data and the features of the historical weather data at different time steps. The multi-head attention mechanism can focus on different feature spaces. Then, through a feedforward neural network, deep learning of feature attributes is performed to establish an energy prediction model, and finally the energy data prediction value is obtained.

[0078] For example, the energy and weather data series is X y =[X y (t1),X y (t2),…,X y (t i ),…,X y (t T )],X y (t i )=[P V (t i ),P L (t i ),W(t i )], where 1≤i≤T, P V (t i ) is t i The historical active power data of distributed energy at the moment, P L (t i ) is t i The historical load power data of distributed energy at the moment, W(t i ) is t i Historical weather data at any time.

[0079] Furthermore, the multi-head self-attention mechanism is used to calculate the dependency between the energy and weather data sequences to obtain the multi-head self-attention score, including:

[0080] adding position codes to the energy and weather data sequences according to time information of the energy and weather data sequences;

[0081] A multi-head self-attention mechanism is adopted to calculate the dependency of the energy and weather data sequences with the added position encoding, and obtain a multi-head self-attention score.

[0082] For example, by Add position coding to the energy and weather data sequence, where Z(t i ) is t with positional encoding addedi Energy and weather data series at each moment; X y (t i ) is t i Energy and weather data sequence at the time; PE(t i ) is t i The position encoding of the moment; i is the dimension index, and d is the dimension of the input feature.

[0083] For each time step, by MultiHead(t i )=Concat(head1,head2,…,head h )·W° calculates the multi-head self-attention score, where Concat() is a string concatenation function, there are h independent self-attention heads, and W° is a trainable projection matrix. It should be noted that each attention head is obtained through Calculate the self-attention score of a single head, where the Softmax() function is an activation function for multi-classification tasks, which maps the elements in a vector to the [0,1] interval, and the sum of all elements is 1; Q(t i ) is the time step t i Query vector, representing the query for information at other time steps; K(t i ) T is the time step t i The transpose of the Key vector represents the information of other time steps in the input sequence; V(t i ) is the time step t i The Value vector represents the actual information that needs to be weighted averaged; d k is a scaling factor used to prevent the dot product result from being too large.

[0084] After calculating the multi-head self-attention score at each time step, the feedforward neural network FFT (t i )=ReLU(W1·Multihead(t i )+b1)·W2+b2 constructs the energy load forecasting model. Among them, FFT(t i ) is a feedforward neural network; ReLU() is an activation function in deep learning, which sets the negative part of the input to 0 and keeps the positive part unchanged; W1 is the weight matrix of the first layer of the feedforward neural network; W2 is the weight matrix of the first layer of the feedforward neural network; b1 is the first bias vector in the feedforward neural network; b2 is the second bias vector in the feedforward neural network.

[0085] The energy load forecasting model constructed by the embodiment of the present invention is: in, is the output result of the energy load forecasting model, is the predicted distributed photovoltaic active power, Predicted load power, Linear() is a linear function, FFN(t i ) is the output of the feedforward neural network.

[0086] As a preferred implementation, step S3, fusing the historical energy storage data to calculate the predicted value of energy storage data of the new distribution network, includes:

[0087] Performing fusion processing on the historical energy storage data to obtain fused energy storage data;

[0088] Establishing an energy storage prediction model based on the fused energy storage data;

[0089] Solve the energy storage prediction model and calculate the energy storage data prediction value of the new distribution network.

[0090] Preferably, for multimodal input information, information of different modalities is integrated through a fusion layer. Specifically, through Z C (t i )=Fusion(X C (t i )),X C (t i )=[P CC (t i ),P CF (t i ),S C (t i )] to obtain fusion energy storage data; where P CC (t i ) is t i The historical charging power of distributed energy storage at the time, P CF (t i ) is t i The historical discharge power of distributed energy storage at the time, S C (t i ) is t i The historical SOC data of distributed energy storage at the moment; Fusion() is the fusion operation function of the fusion layer.

[0091] Furthermore, establishing an energy storage prediction model based on the fused energy storage data includes:

[0092] Adopting a multimodal AI-based algorithm, and establishing an energy storage charging power prediction model and an energy storage discharging power prediction model respectively according to the fused energy storage data;

[0093] An energy storage prediction model is obtained according to the energy storage charging power prediction model and the energy storage discharging power prediction model.

[0094] Furthermore, the multimodal AI algorithm is used to establish an energy storage charging power prediction model and an energy storage discharging power prediction model according to the fused energy storage data, including:

[0095] pass Establish a storage charging power prediction model; where Z C (t) is the fusion energy storage data at time t, f C (Z C (t)) is the energy storage charging power prediction model built based on the multimodal AI algorithm. is the distributed energy storage charging power calculated based on the fused energy storage data at time t;

[0096] pass Establish a storage charging power prediction model; where Z C (t) is the fusion energy storage data at time t, f F (Z C (t)) is the energy storage discharge power prediction model built based on the multimodal AI algorithm. is the distributed energy storage discharge power calculated based on the fused energy storage data at time t.

[0097] In the above scheme, by fusing the historical energy storage data, it is possible to mine the correlation between various historical energy storage data such as historical SOC data of distributed energy storage, historical charging power data and historical discharge power data, and further accurately predict future energy storage data.

[0098] As a preferred implementation, step S4, constructing a new distribution network source-grid-load-storage integrated operation model according to the historical distribution network price data, the energy data prediction value and the energy storage data prediction value, includes:

[0099] Obtaining the total operating cost of the new distribution network according to the historical distribution network price data, the energy data prediction value and the energy storage data prediction value;

[0100] Obtaining a comprehensive power grid response deviation of the new distribution network according to the energy data prediction value and the energy storage data prediction value;

[0101] generating an objective function according to the total operating cost and the comprehensive grid response deviation;

[0102] Establishing constraints of the objective function;

[0103] According to the objective function and the constraints, a new distribution network source-grid-load-storage integrated operation model is constructed.

[0104] In some embodiments, by The total operating cost of the new distribution network is obtained; where T is the cycle, C S (t) is the operating cost of distributed energy storage at time t, C V (t) is the operating cost of distributed energy at time t, C L (t) is the operating cost of the load at time t; is the predicted value of energy storage charging power at time t; is the predicted value of energy storage discharge power at time t; is the load power forecast value at time t; G is the historical electricity price; C is the energy storage discharge price; V is the historical photovoltaic grid-connected price.

[0105] pass Calculate the comprehensive grid response deviation of the new distribution network; where P Z (t) is the comprehensive response power at time t.

[0106] According to the energy data prediction value and the energy storage data prediction value, the generated objective function is: Among them, λ is the preset deviation factor.

[0107] Furthermore, the constraint conditions for establishing the objective function include:

[0108] According to the upper and lower limit constraints of the distributed energy storage charging power, a first constraint condition is established;

[0109] According to the upper and lower limit constraints of the distributed energy storage discharge power, a second constraint condition is established;

[0110] According to the upper and lower limits of distributed energy photovoltaic power, the third constraint condition is established;

[0111] According to the upper and lower limit constraints of distributed energy load power, the fourth constraint condition is established;

[0112] The constraint conditions of the objective function are established according to the first constraint condition, the second constraint condition, the third constraint condition and the fourth constraint condition.

[0113] In some embodiments, the first constraint The second constraint is The third constraint condition is The fourth constraint condition is Among them, P CCmin is the lower limit of the distributed energy storage charging power, P CC max is the upper limit of the distributed energy storage charging power; P CF min is the lower limit of the distributed energy storage discharge power, P CF max is the upper limit of the distributed energy storage discharge power; P L min is the lower limit of distributed energy photovoltaic power, P L max is the upper limit of distributed energy photovoltaic power; P V min is the lower limit of distributed energy load power, P V max is the upper limit of distributed energy load power.

[0114] It should be noted that when solving the running model, an artificial intelligence model, such as a GPT model, can be used. By converting the GPT model into an optimization solver, a solution that meets the objective function and constraints can be calculated and solved.

[0115] As a preferred implementation manner, step S5, solving the operation model to obtain a power allocation strategy, and scheduling the new distribution network according to the power allocation strategy, includes:

[0116] Establishing a loss function of the running model;

[0117] The operation model is solved with the goal of minimizing the loss function to obtain a power allocation strategy; in the process of solving the operation model, the learnable parameters in the solution model are updated at each time step;

[0118] The new power distribution network is dispatched according to the power allocation strategy.

[0119] Exemplarily, the loss function is in, is the input vector to solve the model at time t; is the distributed energy storage charging power predicted at time t; is the predicted distributed energy storage discharge power at time t; is the distributed photovoltaic power predicted at time t; is the load power predicted at time t; C total is the total operating cost of the new distribution network including source, grid, load and storage; ΔP is the integrated comprehensive response grid deviation; λ is the deviation.

[0120] It should be noted that when solving the running model, the input of the running model is iterated according to the solution conditions to finally generate a set of vectors, that is, The optimal solution of .

[0121] Furthermore, the operation model is solved with the goal of minimizing the loss function to obtain a power allocation strategy; in the process of solving the operation model, the learnable parameters in the solution model are updated at each time step, including:

[0122] According to the loss function, a reward function is obtained;

[0123] Calculating the expectation of the reward function and the learning rate of the running model, and optimizing the learnable parameters in the running model at each time step;

[0124] Based on the continuously optimized learnable parameters and with the goal of minimizing the loss function, the operation model is solved to obtain a power allocation strategy.

[0125] Exemplarily, the reward function obtained according to the loss function is:

[0126]

[0127] At each time step, the policy gradient method is used to update the learnable parameter θ in the running model. Specifically, Optimize the learnable parameters in the solution model, where η is the learning rate; is the gradient descent function; E[R(t)] is the expectation of the reward function. In an embodiment of the present invention, the solution model is a GPT model.

[0128] In the above scheme, through multiple iterations, the solution model will gradually learn to generate the optimal solution that meets the objective function and constraints, and by continuously updating the model parameters, it will eventually obtain a set of optimal distributed energy storage power, distributed photovoltaic power and load power as a power allocation strategy.

[0129] A new type of distribution network coordinated dispatching method provided by an embodiment of the present invention can coordinately dispatch distributed energy storage power, distributed energy power and load power, give full play to the potential of distributed energy and distributed energy storage in the new distribution network, and realize the integrated operation of source, grid, load and storage of the new distribution network.

[0130] See also Figure 2 , an embodiment of the present invention provides a schematic diagram of the structure of a new type of distribution network collaborative dispatching system. The new type of distribution network collaborative dispatching system includes a data acquisition module 11, an energy data prediction module 12, an energy storage data prediction module 13, a model building module 14 and a strategy generation module 15, wherein:

[0131] The data acquisition module 11 is used to collect historical source-grid-load-storage data of the new distribution network; the historical source-grid-load-storage data includes historical energy data, historical distribution network price data, historical weather data and historical energy storage data;

[0132] An energy data prediction module 12, configured to calculate an energy data prediction value of the new distribution network based on the historical energy data and the historical weather data;

[0133] The energy storage data prediction module 13 is used to perform fusion processing on the historical energy storage data and calculate the energy storage data prediction value of the new distribution network;

[0134] A model building module 14 is used to build a new distribution network source-grid-load-storage integrated operation model according to the historical distribution network price data, the energy data prediction value and the energy storage data prediction value;

[0135] The strategy generation module 15 is used to solve the operation model, obtain the power allocation strategy, and dispatch the new distribution network according to the power allocation strategy; the power allocation strategy includes distributed energy storage power, distributed energy power and load power.

[0136] As a preferred implementation, the energy data prediction module 12 includes:

[0137] A sequence generating unit, used for obtaining energy and weather data sequences according to the historical energy data and the historical weather data;

[0138] A self-attention score calculation unit, used to calculate the dependency between the energy and weather data sequences using a multi-head self-attention mechanism to obtain a multi-head self-attention score;

[0139] An energy prediction model building unit, used to build an energy prediction model according to the multi-head self-attention score through a feedforward neural network;

[0140] The energy prediction model solving unit is used to solve the energy prediction model and calculate the energy data prediction value of the new distribution network.

[0141] Further, preferably, the self-attention score calculation unit is specifically used to:

[0142] adding position codes to the energy and weather data sequences according to time information of the energy and weather data sequences;

[0143] A multi-head self-attention mechanism is adopted to calculate the dependency of the energy and weather data sequences with the added position encoding, and obtain a multi-head self-attention score.

[0144] As a preferred implementation, the energy storage data prediction module 13 includes:

[0145] A historical energy storage data fusion unit, used for fusing the historical energy storage data to obtain fused energy storage data;

[0146] An energy storage prediction model establishing unit, used to establish an energy storage prediction model according to the fused energy storage data;

[0147] The energy storage prediction model solving unit is used to solve the energy storage prediction model and calculate the energy storage data prediction value of the new distribution network.

[0148] Furthermore, the energy storage prediction model building unit is used to:

[0149] Adopting a multimodal AI-based algorithm, and establishing an energy storage charging power prediction model and an energy storage discharging power prediction model respectively according to the fused energy storage data;

[0150] An energy storage prediction model is obtained according to the energy storage charging power prediction model and the energy storage discharging power prediction model.

[0151] Furthermore, the multimodal AI algorithm is used to establish an energy storage charging power prediction model and an energy storage discharging power prediction model according to the fused energy storage data, including:

[0152] pass Establish a storage charging power prediction model; where Z C (t) is the fusion energy storage data at time t, f C (Z C (t)) is the energy storage charging power prediction model built based on the multimodal AI algorithm. is the distributed energy storage charging power calculated based on the fused energy storage data at time t;

[0153] pass Establish a storage charging power prediction model; where Z C (t) is the fusion energy storage data at time t, f F (Z C (t)) is the energy storage discharge power prediction model built based on the multimodal AI algorithm. is the distributed energy storage discharge power calculated based on the fused energy storage data at time t.

[0154] As a preferred implementation, the model building module 14 includes:

[0155] A total operation cost calculation unit, used to obtain the total operation cost of the new distribution network according to the historical distribution network price data, the energy data prediction value and the energy storage data prediction value;

[0156] A comprehensive power grid response deviation calculation unit, used to obtain a comprehensive power grid response deviation of the new distribution network according to the energy data prediction value and the energy storage data prediction value;

[0157] An objective function generating unit, configured to generate an objective function according to the total operating cost and the comprehensive power grid response deviation;

[0158] A constraint condition establishing unit, used to establish the constraint condition of the objective function;

[0159] The operation model establishment unit is used to construct an operation model of a new distribution network integrating source, grid, load and storage according to the objective function and the constraint conditions.

[0160] Furthermore, the constraint condition establishing unit is specifically used to:

[0161] According to the upper and lower limit constraints of the distributed energy storage charging power, a first constraint condition is established;

[0162] According to the upper and lower limit constraints of the distributed energy storage discharge power, a second constraint condition is established;

[0163] According to the upper and lower limits of distributed energy photovoltaic power, the third constraint condition is established;

[0164] According to the upper and lower limit constraints of distributed energy load power, the fourth constraint condition is established;

[0165] The constraint conditions of the objective function are established according to the first constraint condition, the second constraint condition, the third constraint condition and the fourth constraint condition.

[0166] As a preferred implementation, the strategy generation module 15 includes:

[0167] A loss function establishing unit, used to establish a loss function of the running model;

[0168] An operation model solving unit is used to solve the operation model with the goal of minimizing the loss function to obtain a power allocation strategy; in the process of solving the operation model, the learnable parameters in the solution model are updated at each time step;

[0169] A power grid dispatching unit is used to dispatch the new power distribution network according to the power allocation strategy.

[0170] A new type of distribution network collaborative dispatching system provided by an embodiment of the present invention can coordinate the dispatching of distributed energy storage power, distributed energy power and load power, give full play to the potential of distributed energy and distributed energy storage in the new type of distribution network, and realize the integrated operation of source, grid, load and storage of the new distribution network.

[0171] A person skilled in the art can understand that all or part of the processes in the above-mentioned embodiments can be implemented by instructing the relevant hardware through a computer program, and the program can be stored in a computer-readable storage medium, and when the program is executed, it can include the processes of the embodiments of the above-mentioned methods. The storage medium can be a disk, an optical disk, a read-only memory (ROM) or a random access memory (RAM), etc.

[0172] The above is a preferred embodiment of the present invention. It should be pointed out that a person skilled in the art can make several improvements and modifications without departing from the principle of the present invention. These improvements and modifications are also considered to be within the scope of protection of the present invention.

Claims

1. A new coordinated dispatching method for distribution network, characterized in that: include: Collecting historical source, grid, load and storage data of the new distribution network; the historical source, grid, load and storage data includes historical energy data, historical distribution network price data, historical weather data and historical energy storage data; Calculate the energy data forecast value and the load data forecast value of the new distribution network according to the historical energy data and the historical weather data; Performing fusion processing on the historical energy storage data to calculate and obtain a predicted value of energy storage data of the new distribution network; According to the historical distribution network price data, the energy data forecast value, the load data forecast value and the energy storage data forecast value, a new distribution network source-grid-load-storage integrated operation model is constructed; The operation model is solved to obtain a power allocation strategy, and the new distribution network is dispatched according to the power allocation strategy; the power allocation strategy includes distributed energy storage power, distributed photovoltaic power and load power.

2. A new type of coordinated dispatching method for distribution network according to claim 1, characterized in that: The step of calculating the energy data forecast value and the load data forecast value of the new distribution network according to the historical energy data and the historical weather data includes: Obtaining energy and weather data sequences according to the historical energy data and the historical weather data; A multi-head self-attention mechanism is used to calculate the dependency between the energy and weather data sequences to obtain a multi-head self-attention score; Establishing an energy load forecasting model based on the multi-head self-attention scores through a feedforward neural network; Solve the energy load forecasting model and calculate the energy data forecast value and load data forecast value of the new distribution network.

3. A new type of coordinated dispatching method for distribution network as claimed in claim 2, characterized in that: The multi-head self-attention mechanism is used to calculate the dependency between the energy and weather data sequences to obtain the multi-head self-attention score, including: adding position codes to the energy and weather data sequences according to time information of the energy and weather data sequences; A multi-head self-attention mechanism is adopted to calculate the dependency of the energy and weather data sequences with the added position encoding, and obtain a multi-head self-attention score.

4. A new type of coordinated dispatching method for distribution network according to claim 1, characterized in that: The fusing and processing the historical energy storage data to calculate the predicted value of energy storage data of the new distribution network includes: Performing fusion processing on the historical energy storage data to obtain fused energy storage data; Establishing an energy storage prediction model based on the fused energy storage data; Solve the energy storage prediction model and calculate the energy storage data prediction value of the new distribution network.

5. A new type of coordinated dispatching method for distribution network as claimed in claim 4, characterized in that: The step of establishing an energy storage prediction model based on the fused energy storage data includes: Adopting a multimodal AI-based algorithm, and establishing an energy storage charging power prediction model and an energy storage discharging power prediction model respectively according to the fused energy storage data; An energy storage prediction model is obtained according to the energy storage charging power prediction model and the energy storage discharging power prediction model.

6. A new type of coordinated dispatching method for distribution network as claimed in claim 5, characterized in that: The multimodal AI algorithm is used to establish an energy storage charging power prediction model and an energy storage discharging power prediction model according to the fused energy storage data, including: pass Establish a storage charging power prediction model; where Z C (t) is the fusion energy storage data at time t, f C (Z C (t)) is the energy storage charging power prediction model built based on the multimodal AI algorithm. is the distributed energy storage charging power calculated based on the fused energy storage data at time t; pass Establish a storage charging power prediction model; where Z C (t) is the fusion energy storage data at time t, f F (Z C (t)) is the energy storage discharge power prediction model built based on the multimodal AI algorithm. is the distributed energy storage discharge power calculated based on the fused energy storage data at time t.

7. A new type of coordinated dispatching method for distribution network according to claim 1, characterized in that: The constructing of a new distribution network source-grid-load-storage integrated operation model according to the historical distribution network price data, the energy data forecast value, the load data forecast value and the energy storage data forecast value comprises: Obtaining the total operating cost of the new distribution network according to the historical distribution network price data, the energy data prediction value and the energy storage data prediction value; Obtaining a comprehensive power grid response deviation of the new distribution network according to the energy data prediction value, the load data prediction value and the energy storage data prediction value; generating an objective function according to the total operating cost and the comprehensive grid response deviation; Establishing constraints of the objective function; According to the objective function and the constraints, a new distribution network source-grid-load-storage integrated operation model is constructed.

8. A new type of coordinated dispatching method for distribution network as claimed in claim 7, characterized in that: The constraint conditions for establishing the objective function include: According to the upper and lower limit constraints of the distributed energy storage charging power, a first constraint condition is established; According to the upper and lower limit constraints of the distributed energy storage discharge power, a second constraint condition is established; According to the upper and lower limits of distributed energy photovoltaic power, the third constraint condition is established; According to the upper and lower limit constraints of distributed energy load power, the fourth constraint condition is established; The constraint conditions of the objective function are established according to the first constraint condition, the second constraint condition, the third constraint condition and the fourth constraint condition.

9. A new type of coordinated dispatching method for distribution network according to claim 1, characterized in that: The step of solving the operation model to obtain a power allocation strategy, and scheduling the new distribution network according to the power allocation strategy, includes: Establishing a loss function of the running model; The operation model is solved with the goal of minimizing the loss function to obtain a power allocation strategy; in the process of solving the operation model, the learnable parameters in the solution model are updated at each time step; The new power distribution network is dispatched according to the power allocation strategy.

10. A new type of coordinated dispatching method for distribution network according to claim 9, characterized in that: The operation model is solved with the goal of minimizing the loss function to obtain a power allocation strategy; In the process of solving the operation model, updating the learnable parameters in the operation model at each time step includes: According to the loss function, a reward function is obtained; Calculating the expectation of the reward function and the learning rate of the running model, and optimizing the learnable parameters in the running model at each time step; Based on the continuously optimized learnable parameters and with the goal of minimizing the loss function, the operation model is solved to obtain a power allocation strategy.

11. A new type of coordinated dispatching system for distribution network, characterized in that: include: A data acquisition module is used to collect historical source-grid-load-storage data of the new distribution network; the historical source-grid-load-storage data includes historical energy data, historical distribution network price data, historical weather data and historical energy storage data; An energy data prediction module, used to calculate the energy data prediction value and the load data prediction value of the new distribution network according to the historical energy data and the historical weather data; An energy storage data prediction module is used to perform fusion processing on the historical energy storage data and calculate the energy storage data prediction value of the new distribution network; A model building module, used to build a new distribution network source-grid-load-storage integrated operation model according to the historical distribution network price data, the energy data forecast value, the load data forecast value and the energy storage data forecast value; A strategy generation module is used to solve the operation model, obtain a power allocation strategy, and dispatch the new distribution network according to the power allocation strategy; the power allocation strategy includes distributed energy storage power, distributed energy power and load power.

Images