Power electronic power distribution network energy optimization method

Abstract

The invention discloses a power electronic power distribution network energy optimization method which mainly solves the problems that an existing power distribution network energy optimization mode is large in calculation amount and complex in solution. The optimization method comprises the following steps: S1, determining a receiving-end power grid which is accessed into a power distribution network and is used for energy optimization, and establishing a special communication channel for energy optimization between the receiving-end power grid and a power utilization party; s2, submitting a historical annual power utilization load, a historical time-sharing power utilization period load and an annual predicted power utilization load by a power utilization party through an energy optimization dedicated communication channel; s3, the power distribution network establishes an energy optimization model of the receiving end power grid based on the demand response of the power consumer; and S4, establishing a constraint condition of the energy optimization model based on load control, and solving the energy optimization model with a receiving end power grid benefit maximization target. According to the method, model training is carried out on the data by adopting the convolutional neural prediction network model through the historical year power consumption load and the historical time-sharing power consumption period load, the training model is simple and has few parameters, convergence is faster during network training, obtained prediction data are more accurate, and accurate energy optimization management of the receiving end power grid can be better achieved.

Description

Technical Field

[0001] The present invention belongs to the technical field of power distribution networks, and in particular, relates to an energy optimization method for a power electronic power distribution network. Background Art

[0002] The power grid is an important technological product and the most representative symbol of human civilization's transition from the mechanical age to the electrical age. The power grid usually refers to a large-scale system that connects power generation and power consumption, and is managed uniformly by transmission equipment, substation equipment, distribution equipment, and corresponding secondary systems to provide users with a safe and stable source of power.

[0003] Energy Management Optimization (EMO), or Demand Response Management (DRM), is one of the main features of smart grids. Energy Management Optimization refers to management activities that are carried out under the support of national policies and regulations, through the joint participation of power supply companies and various users, and effective incentive and guidance measures to improve the power efficiency of users and change their power habits while meeting their power needs, ultimately improving the energy efficiency of the power grid system, enhancing the reliability and effectiveness of the distribution grid, reducing the environmental impact of the power production process, and achieving the best social benefits, benefits for all parties, and cost reduction.

[0004] In the existing distribution network energy optimization management, the algorithm for predicting the annual power load of power users adopts the random sample extraction method based on the Monte Carlo method. However, this method is extremely computationally intensive and requires solving a large number of optimization problems. When facing multi-object optimization scheduling tasks, the relevant optimization scheduling methods have weak adjustment capabilities and usually only consider active power or reactive power, without considering both active power and reactive power at the same time, so that the optimization capability is not perfect. Summary of the invention

[0005] The purpose of the present invention is to provide a method for optimizing the energy of a power electronic distribution network, mainly to solve the problem that the existing energy optimization methods for distribution networks have large amount of calculation and complex solutions.

[0006] To achieve the above purpose, the technical solution adopted by the present invention is as follows:

[0007] A method for optimizing energy in a power electronic distribution network comprises the following steps:

[0008] S1, determine the receiving power grid connected to the distribution network for energy optimization, and establish a dedicated energy optimization communication channel between the receiving power grid and the power user;

[0009] S2, the electricity user submits the historical annual electricity load, the historical time-sharing electricity load and the annual predicted electricity load through the energy optimization dedicated communication channel;

[0010] S3, the distribution network establishes an energy optimization model for the receiving power grid based on the demand response of the power user;

[0011] S4, based on load control, establish the constraint conditions of the energy optimization model, and solve the energy optimization model with the goal of maximizing the benefit of the receiving power grid.

[0012] Furthermore, in step S1, the method for establishing the energy optimization dedicated communication channel is as follows:

[0013] S11, obtaining the communication ID of the receiving power grid connected to the distribution network for energy optimization;

[0014] S12, establishing a communication area for the power user for energy optimization, with the distance between the power user and the substation as the diameter and the midpoint of the line segment between the power user and the substation as the center;

[0015] S13, determining whether there is a communication relay node of the receiving-end power grid in the communication area of ​​the electricity user;

[0016] If it does not exist, the electricity user directly communicates with the substation, and the substation communicates with the receiving power grid to complete the establishment of the energy optimization dedicated communication channel; if it does exist, proceed to step S4;

[0017] S14, taking the distance between the electricity user and the substation as the diameter and the midpoint of the line segment between the electricity user and the substation as the center of the circle, set the sub-area that contains the communication relay node and is concentric with the electricity user's communication area as the communication relay area; the electricity user communicates with the receiving power grid through the relay node to complete the establishment of a dedicated communication channel for energy optimization.

[0018] Furthermore, the annual predicted electricity load is obtained by training a convolutional neural prediction network model; the following steps are included:

[0019] S21, define the historical annual electricity load data characteristic input vector F = [f1, f2, ..., f n ], then the output of the convolutional neural prediction network model is: h = [h1, h2, ..., h k ]∈G (N-S+1 ) K ,in:

[0020] h k =G(conv(w k *f+a k ))

[0021] Among them, the convolutional layer of the convolutional neural prediction network model has K convolution kernels, the size is 1*S, and the coefficient is w k ∈G S,k=1,2,…,K; conv represents convolution transformation, a k is the bias of the convolution kernel, G is the nonlinear activation function Leaky ReLu;

[0022] S2, assuming that the convolutional neural network model has L layers, where L M is the fully connected layer, L L As the output layer, the number of input samples is n, and the annual predicted electricity load is calculated as follows:

[0023] h l =C(w l F+a l ), L M ≤l <L

[0024]

[0025] y L =w l h l +a L

[0026]

[0027] In the formula, h l is the output of the lth layer, w l and a L is the parameter learned by the network, y L is the value of the output layer that is not activated, X is the annual predicted power load output value, s t The output of each sample corresponds to the target.

[0028] Furthermore, in step S3, the energy optimization model is:

[0029] minJ=J m +J v

[0030]

[0031] In the formula, J m is the cost function of annual power load forecast, J v Cost function for predicting the annual power generation cost of the receiving power grid; Q, R are self-defined positive definite matrices, i is the time pointer, t is the current time, and N is the length of the prediction time domain; is the expected operator; x i is the input variable, u i To control the amount.

[0032] Furthermore, in step S4, the established constraints include:

[0033] Time-of-use electricity price response constraints:

[0034]

[0035] ∑P high,t represents the sum of high-price power during high-demand periods, ∑P low,t The sum of low-price power during periods of low demand; is the total power generated by the receiving power grid;

[0036] Real-time electricity price constraints:

[0037] E 用,t = E 发,t

[0038] Among them, E 用,t represents the real-time electricity price, E 发,t Indicates the real-time power generation cost price;

[0039] Peak price response:

[0040] E 尖峰,t >E 其他,t

[0041] Among them, E 尖峰,t represents the peak period electricity price, E 其他,t Indicates electricity prices for other periods of time.

[0042] Compared with the prior art, the present invention has the following beneficial effects:

[0043] (1) The present invention establishes a dedicated communication channel between the receiving power grid and the power user in the energy optimization of the distribution network. Through the dedicated communication channel, the receiving power grid can quickly and accurately obtain the power consumption data of the power user in real time, so as to realize energy management through the receiving power grid quickly and accurately and optimize the power generation power of the receiving power grid.

[0044] (2) The present invention uses a convolutional neural prediction network model to train the data based on the historical annual electricity load and the historical time-sharing electricity load. The training model is simple and has few parameters, so the convergence is faster when training the network, and the predicted data obtained is more accurate, which is more conducive to the precise energy optimization management of the receiving power grid. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] Figure 1 It is a schematic diagram of the structure of the present invention.

[0046] Figure 2 Schematic diagram of the prediction results of the convolutional neural prediction network model in an embodiment of the present invention. DETAILED DESCRIPTION

[0047] The present invention is further described below in conjunction with the accompanying drawings and embodiments. The embodiments of the present invention include but are not limited to the following embodiments.

[0048] like Figure 1 As shown, the present invention discloses a method for optimizing the energy of a power electronic distribution network. The important technical prerequisite for the implementation of this method is the widespread popularization of smart meters and the realization of two-way communication between the power distributor and the power user. The implementation of energy optimization management of the smart distribution network requires the joint participation of the power distributor and the power user. In this embodiment, first, a receiving power grid connected to the distribution network for energy optimization is determined, and a dedicated communication channel for energy optimization is established between the receiving power grid and the power user; the method for establishing the dedicated communication channel is as follows: the first step is to obtain the communication ID of the receiving power grid connected to the distribution network for energy optimization; the second step is to establish a communication area for the power user for energy optimization with the distance between the power user and the substation as the diameter and the midpoint of the line segment between the power user and the substation as the center; the third step is to determine whether there is a communication relay node of the receiving power grid in the power user's communication area; if not, the power user directly communicates with the substation, and the substation communicates with the receiving power grid to complete the establishment of the dedicated communication channel for energy optimization; if so, proceed to the fourth step; the fourth step is to set the sub-area containing the communication relay node and forming a concentric circle with the power user's communication area as the communication relay area with the distance between the power user and the substation as the diameter and the midpoint of the line segment between the power user and the substation as the center; the power user communicates with the receiving power grid through the relay node to complete the establishment of the dedicated communication channel for energy optimization. Through this dedicated communication channel, the receiving power grid can quickly and accurately obtain the power consumption data of the power user in real time, which is convenient for quickly and accurately realizing energy management through the receiving power grid and optimizing the power generation power of the receiving power grid.

[0049] Subsequently, the electricity user submits the historical annual electricity load, the historical time-sharing electricity load and the annual predicted electricity load through the energy optimization dedicated communication channel; the annual predicted electricity load is obtained by training the convolutional neural prediction network model: First, define the historical annual electricity load data feature input vector F = [f1, f2, ..., f n ], then the output of the convolutional neural prediction network model is: h = [h1, h2, ..., h k ]∈G (N-S+1)K ,in:

[0050] h k =G(conv(w k *f+a k ))

[0051] Among them, the convolutional layer of the convolutional neural prediction network model has K convolution kernels, the size is 1*S, and the coefficient is w k ∈G S ,k=1,2,…,K; conv represents convolution transformation, ak is the bias of the convolution kernel, G is the nonlinear activation function Leaky ReLu; suppose the convolutional neural prediction network model has L layers, where L M is the fully connected layer, L L As the output layer, the number of input samples is n, and the annual predicted electricity load is calculated as follows:

[0052] h l =C(w l F+a l ),L M ≤l <L

[0053]

[0054] y L =w l h l +a L

[0055]

[0056] In the formula, h l is the output of the lth layer, w l and a L is the parameter learned by the network, y L is the value of the output layer that is not activated, X is the annual predicted power load output value, s t The output of each sample corresponds to the target. In this embodiment, the historical load of the previous three years is selected as the initial input training data convolutional neural prediction network model, and the prediction results are displayed in a 24-hour cycle. The prediction results are as follows: Figure 2 shown.

[0057] Then the distribution network establishes the energy optimization model of the receiving power grid based on the demand response of the power user:

[0058] minJ=J m +J v

[0059]

[0060] In the formula, J m is the cost function of annual power load forecast, J v Cost function for predicting the annual power generation cost of the receiving power grid; Q, R are self-defined positive definite matrices, i is the time pointer, t is the current time, and N is the length of the prediction time domain; is the expected operator; x i is the input variable, u i To control the amount.

[0061] Finally, based on load control, the constraints of the energy optimization model are established, and the energy optimization model is solved with the goal of maximizing the benefit of the receiving power grid. The established constraints include:

[0062] Time-of-use electricity price response constraints:

[0063]

[0064] ∑P high,t represents the sum of high-price power during high-demand periods, ∑P low,t The sum of low-price power during periods of low demand; is the total power generated by the receiving power grid;

[0065] Real-time electricity price constraints:

[0066] E 用,t =E 发,t

[0067] Among them, E 用,t represents the real-time electricity price, E 发,t Indicates the real-time power generation cost price;

[0068] Peak price response:

[0069] E 尖峰,t >E 其他,t

[0070] Among them, E 尖峰,t represents the peak period electricity price, E 其他,t Indicates electricity prices for other periods of time.

[0071] Through the above design, the present invention establishes a dedicated communication channel between the receiving power grid and the electricity user in the energy optimization of the distribution network. Through the dedicated communication channel, the receiving power grid can obtain the electricity consumption data of the electricity user in real time, quickly and accurately, which is convenient for quickly and accurately realizing energy management through the receiving power grid and optimizing the power generation power of the receiving power grid.

[0072] The above embodiment is only one of the preferred implementation modes of the present invention and should not be used to limit the protection scope of the present invention. Any changes or modifications that are made to the main design concept and spirit of the present invention and have no substantive significance, and the technical problems they solve are still consistent with the present invention, should be included in the protection scope of the present invention.

Claims

1. A method for optimizing energy in a power electronic distribution network, characterized in that: The following steps are involved: S1, determine the receiving power grid connected to the distribution network for energy optimization, and establish a dedicated energy optimization communication channel between the receiving power grid and the power user; S2, the electricity user submits the historical annual electricity load, the historical time-sharing electricity load and the annual predicted electricity load through the energy optimization dedicated communication channel; S3, the distribution network establishes an energy optimization model for the receiving power grid based on the demand response of the power user; S4, based on load control, establish the constraint conditions of the energy optimization model, and solve the energy optimization model with the goal of maximizing the benefit of the receiving power grid.

2. A method for optimizing energy in a power electronic distribution network according to claim 1, characterized in that: In step S1, the method for establishing the energy optimization dedicated communication channel is as follows: S11, obtaining the communication ID of the receiving power grid connected to the distribution network for energy optimization; S12, establishing a communication area for the power user for energy optimization, with the distance between the power user and the substation as the diameter and the midpoint of the line segment between the power user and the substation as the center; S13, determining whether there is a communication relay node of the receiving-end power grid in the communication area of ​​the electricity user; If it does not exist, the electricity user directly communicates with the substation, and the substation communicates with the receiving power grid to complete the establishment of the energy optimization dedicated communication channel; if it does exist, proceed to step S4; S14, taking the distance between the electricity user and the substation as the diameter and the midpoint of the line segment between the electricity user and the substation as the center of the circle, set the sub-area that contains the communication relay node and is concentric with the electricity user's communication area as the communication relay area; the electricity user communicates with the receiving power grid through the relay node to complete the establishment of a dedicated communication channel for energy optimization.

3. The method for optimizing energy of a power electronic distribution network according to claim 2, characterized in that: The annual predicted electricity load is obtained by training a convolutional neural prediction network model; the steps include: S21, define the historical annual electricity load data characteristic input vector F = [f1, f2, ..., f n ], then the output of the convolutional neural prediction network model is: h = [h1, h2, ..., hk] ∈ G (N-S+1)K ,in: h k =G(conv(w k *f+a k )) Among them, the convolutional layer of the convolutional neural prediction network model has K convolution kernels, the size is 1*S, and the coefficient is w k ∈G S , k = 1, 2, ..., K; conv represents convolution transformation, a k is the bias of the convolution kernel, G is the nonlinear activation function Leaky ReLu; S2, assuming that the convolutional neural network model has L layers, where L M is the fully connected layer, L L As the output layer, the number of input samples is n, and the annual predicted electricity load is calculated as follows: h l =C(w l F+a l ),L M ≤l<L y L =w l h l +a L In the formula, h l is the output of the,th layer, w l and a L is the parameter learned by the network, y L is the value of the output layer that is not activated, X is the annual predicted power load output value, s t The output of each sample corresponds to the target.

4. A method for optimizing energy in a power electronic distribution network according to claim 3, characterized in that: In step S3, the energy optimization model is: minJ=J m +J v In the formula, J m is the cost function of annual power load forecast, J v Cost function for predicting the annual power generation cost of the receiving power grid; Q, R are self-defined positive definite matrices, i is the time pointer, t is the current time, and N is the length of the prediction time domain; is the expected operator; x i is the input variable, u i To control the amount.

5. A method for optimizing energy in a power electronic distribution network according to claim 4, characterized in that: In step S4, the established constraints include: Time-of-use electricity price response constraints: ∑P high，t represents the sum of high-price power during high-demand periods, ∑P low，t The sum of low-price power during periods of low demand; is the total power generated by the receiving power grid; Real-time electricity price constraints: AND 用，t =And 发，t Among them, E 用，t represents the real-time electricity price, E 发，t Indicates the real-time power generation cost price; Peak price response: AND 尖峰，t >And 其他，t Among them, E 尖峰，t represents the peak electricity price, E 其他，t Indicates electricity prices for other periods of time.

Images