Power distribution network rapid cooperative scheduling method considering participation of demand side distributed resources

Abstract

The invention discloses a power distribution network rapid cooperative scheduling method considering participation of demand side distributed resources. The method comprises the following steps: step 1, selecting a model construction area, analyzing output characteristics of conventional thermal power generation, distributed wind power generation, distributed photovoltaic power generation and an interruptible load in a transmission-distribution power grid architecture, and constructing a predictive mathematical model; 2, constructing a mathematical model of a power transmission layer in combination with the loss cost of each power generation form in the step 1; 3, constructing a corresponding mathematical model by combining the power utilization and power transmission characteristics of the power transmission layer and the power distribution layer; 4, constructing a dynamic balance loss function of a transmission-distribution power grid architecture, proposing an acceleration penalty term and a convergence iterative algorithm, and continuously training the model until convergence, so that the dynamic response between a power distribution layer and a power transmission layer is more timely; and step 5, applying the model obtained by training to the power distribution network containing the distributed resources. According to the method, the dynamic response speed of power distribution network dispatching can be improved, the overall power supply cost is reduced, and good economic benefits can be brought.

Description

Technical Field

[0001] The present invention relates to an efficient scheduling algorithm for distributed power sources between the distribution layer and the transmission layer of a power grid, and in particular to a fast coordinated scheduling method for a distribution network that takes into account the participation of distributed resources on the demand side. Background Art

[0002] Distributed energy dispatch stems from the challenges of traditional energy supply methods, including instability, waste, and environmental pollution. With the rapid development of renewable energy, distributed energy systems have emerged, and their dispatch technology has become a key tool for addressing these issues. Distributed energy dispatch plays a vital role in promoting energy transition and sustainable development by optimizing energy allocation, improving energy efficiency, and enhancing grid security and stability. It not only reduces carbon emissions and protects the environment, but also improves energy supply reliability, reduces energy costs, and promotes economic growth and social equity. Therefore, the research and application of distributed energy dispatch technology has far-reaching implications for achieving a green, low-carbon energy system.

[0003] Distributed energy hierarchical architecture scheduling algorithms are strategies that decompose complex scheduling tasks into multiple layers for processing, aiming to improve scheduling efficiency, flexibility, and scalability. These algorithms typically combine the advantages of centralized and distributed scheduling, optimizing resource allocation and utilization through hierarchical management. Within a distributed energy hierarchical architecture, scheduling algorithms can be divided into multiple layers, each responsible for different scheduling tasks. These algorithms can employ multi-objective optimization or heuristic algorithms to optimize resource allocation with the goals of minimizing cost, maximizing energy efficiency, or satisfying specific constraints. It is important to note that the specific implementation of a distributed energy hierarchical architecture scheduling algorithm depends on factors such as the specific energy system, scheduling objectives, and constraints. Therefore, in practical applications, they require customization and optimization based on specific circumstances.

[0004] Through research on relevant literature and analysis of the current research status, it is found that the efficient scheduling algorithm for distributed power sources between the "distribution layer and transmission layer" still faces the following two major challenges: 1. Distributed energy sources such as solar energy and wind energy are intermittent and uncertain, resulting in unstable energy output. The data characteristics in different regions and at different times vary greatly, and it is difficult to dynamically construct a mathematical model; 2. The distribution of various power sources in the power grid varies greatly, the relationship between the distribution layer and the transmission layer is complex, and the mathematical model is difficult to abstract. In addition, in complex scheduling relationships, the feature description accuracy of existing algorithms is insufficient and the dynamic convergence speed is low or the convergence is difficult. Summary of the Invention

[0005] To solve the above problems, the present invention proposes a method for rapid coordinated dispatching of distribution networks that takes into account the participation of distributed resources on the demand side. In response to problem 1, this patent, based on data collected at the bureau point, analyzes the output characteristics of thermal power generation, distributed wind power generation, distributed photovoltaic power generation, and interruptible loads and constructs a predictive mathematical model by considering multiple environmental influencing factors such as the climbing / slipping rate of thermal generators, combining the Weibull distribution model, and fitting and estimating the measured value and standard deviation of the average wind speed in the wind farm. In response to problem 2, the patent combines the loss cost, power consumption, and transmission characteristics of various distributed power sources to construct a mathematical model of the transmission layer and the distribution layer, abstractly separating the distribution layer and the transmission layer. At the same time, a dynamic balance loss function for the "transmission-distribution" architecture is proposed, and an accelerated penalty term and convergence iterative algorithm are proposed to make the dynamic response between the distribution layer and the transmission layer more timely.

[0006] To achieve this goal, the present invention considers a method for rapid coordinated dispatching of a distribution network with the participation of distributed resources on the demand side, and the method includes the following steps:

[0007] Step 1. Select a model construction area, analyze the output characteristics of conventional thermal power generation, distributed wind power generation, distributed photovoltaic power generation, and interruptible loads in the "transmission-distribution" power grid architecture, and construct a predictive mathematical model;

[0008] Step 2: Build a mathematical model for the transmission layer to ensure grid power demand while maximizing economic benefits, i.e., minimizing total operating costs.

[0009] Step 3. Construct a mathematical model of the distribution layer. The cost of using the kth distribution layer Mainly from the following aspects: the cost of power transmission in the distribution network Energy storage charging and discharging costs Interruptible charge control cost Power generation cost of generator sets Therefore, in order to improve economic efficiency, it is necessary to minimize The objective function of optimization is:

[0010]

[0011] The specific mathematical expression is as follows:

[0012]

[0013] in represents the power received by the distribution layer at time t, and Represent the discharge / charge cost coefficient, and Respectively represent the energy storage discharge / charging power, ρ cntl and denote the control cost coefficient of the oth interruptible load and the attenuation power at time t, respectively. and They represent the cost coefficient of the jth thermal generator in the kth distribution layer, represents the generator power at time t;

[0014] Step 4. Construct a dynamic balance loss function for the transmission-distribution grid architecture, so that the transmission and distribution networks can meet the demands of both parties while minimizing their own losses. Model this dynamic process:

[0015] Step 5: Apply the trained distributed generation collaborative dispatch model with a differentiated "transmission-distribution" architecture to a distribution network containing distributed resources. This improves the dynamic response speed of distribution network dispatch and the resulting economic benefits.

[0016] Further, step 1 is specifically as follows:

[0017] Step 1.1. A conventional thermal power generation model is constructed based on the ramp rate and ramp rate of the corresponding generator set. The mathematical model is:

[0018]

[0019] in, and represents the upper / lower limit of the output power of the i-th generator set at time t; and They represent the maximum output power, the power at time t, and the minimum output power of the i-th generator set respectively; DR represents the maximum limit range of power that the i-th unit can deliver at time t. i and UR i They represent the maximum ramp rate and ramp rate of the i-th unit respectively. By constructing a mathematical model, the transmission capacity of the generator unit at time t is finally determined, thereby realizing the adaptive adjustment of the model;

[0020] Step 1.2. Construct a mathematical model for wind speed probability density prediction. Use a two-parameter Weibull distribution model and introduce the measured value of the wind field average wind speed and the fitted estimate of the standard deviation to determine the model expression:

[0021]

[0022] Where, and They represent the power provided by the i-th wind turbine at time t and the rated output power, V and V respectively. R are wind speed sequence set and rated wind speed, V ci,tand V co,t They represent the cut-in / cut-out wind speed at time t respectively;

[0023] Step 1.3. Construct a mathematical model for distributed photovoltaic power generation and use the maximum likelihood estimation algorithm for the light intensity Beta to determine the photovoltaic timing characteristics. The photovoltaic output probability density function expression is:

[0024]

[0025] Where, and denote the maximum illumination intensity per unit area of ​​the i-th wind turbine and the illumination intensity at time t, respectively; α and β represent shape coefficients; and Γ() represents the Gamma function;

[0026] Combined with the influence of the light-receiving area of ​​the photovoltaic array, the output power of the photovoltaic array is finally determined The mathematical expression is as follows:

[0027]

[0028] Where A represents the photovoltaic array area, η c,t Represents the photovoltaic conversion efficiency of the photovoltaic array at time t, K c is the threshold function;

[0029] Step 1.4. Construct a mathematical model for interruptible load power generation. For interruptible loads, the regulation range is the widest, and the maximum discharge power of the i-th device is recorded as Then the discharge power of the i-th device at time t is The range is

[0030] Furthermore, the objective function C of the step 2 mathematical model is concretized dis The mathematical expression is:

[0031] minC dis =C pun +C the +C sun +C the,b +C us -C sell

[0032] Where C pun and C us C represents the penalty cost when the power of the distribution network tie line exceeds the limit and the risk of wind curtailment occurs, the Represents the power generation cost of thermal generators, C sun represents the loss cost of distributed photovoltaic power generation, C the,b represents the backup cost of thermal generators, C sell Revenue from electricity sales;

[0033] The specific expressions of each term in the above formula are as follows:

[0034]

[0035] T represents the total operation time of the distribution layer, N J and N K They represent the thermal power generator set and the distribution network set respectively; ρ pun Indicates the penalty coefficient for exceeding the limit of transmission power of tie line, represents the transmission power of the tie line of the jth thermal generator at time t; m j 、h j and s j They represent the cost coefficient of the j-th thermal generator, is the power generation of the jth thermal generator at time t, ξ sun represents the loss coefficient of distributed photovoltaic power generation, and Represent the positive / negative spinning reserve coefficient, and denote the positive / negative spinning reserve cost coefficient, ρ us 、 and They represent the wind power curtailment penalty coefficient, the dispatch output at time t, and the predicted output at time t, respectively. and They are respectively represented as the node electricity price and boundary transmission power of the kth distribution network at time t.

[0036] Further, step 4 is specifically as follows:

[0037] Step 4.1. For the kth distribution network, at time t, the ideal boundary transmission power is Equal to the power received by the distribution layer These two variables are in dynamic adjustment. In order to make the loss function capture and accelerate convergence faster, and Based on the difference, a new e-exponential acceleration penalty term is proposed to improve the response speed of the distribution network. The specific objective function can be expressed as:

[0038]

[0039] Where y k,t and z k,t They represent the difference term penalty factor and the acceleration term penalty factor, respectively. The initial values ​​of the iteration are 0.01 and 0.011. The superscript “—” represents a known term.

[0040] Step 4.2. Similarly, to ensure that the transmission network can quickly obtain input from the distribution network, the transmission network needs to be able to quickly feedback the current distribution network transmission gap. The same principle as step 4.1 is used to construct the corresponding objective function for the transmission network in a collective manner as follows:

[0041]

[0042] Where N q represents the set of transmission networks;

[0043] Step 4.3. Formulate the convergence criterion for model iteration. Considering the consistency of dynamic adjustment of distribution network and transmission network, a new convergence algorithm is proposed. The specific expression is as follows:

[0044]

[0045] Where, represents the objective function value corresponding to the area, ε1 and ε2 represent the iterative convergence coefficients, which are set to 1-e respectively. -2 and e -2 , Item can guarantee and In the case of large differences, the objective function can quickly obtain the difference and converge. Energy Guarantee and In the case of small difference, Supplement to ensure that the objective function can converge quickly at different stages;

[0046] Step 4.4. For those that do not meet the convergence conditions in step 4.3, update the penalty factor, and then loop through steps 4.3 and 4.4 until the model converges. The update criteria are as follows:

[0047]

[0048] Where ψ and χ are constants with values ​​ranging from [1.8, 2.1] and [1, 3], respectively, and are set to 1.9 and 2.0.

[0049] The present invention considers the participation of distributed resources on the demand side in the distribution network rapid coordinated scheduling method. It is based on the analysis of the output characteristics of conventional thermal power generation, distributed wind power generation, distributed photovoltaic power generation and interruptible loads in the "transmission-distribution" architecture and the construction of a predictive mathematical model. Then, the mathematical models of the transmission layer and the distribution layer are constructed respectively. Then, a dynamic balance loss function, an accelerated penalty term and a convergence iterative algorithm are proposed, which makes the dynamic response between the distribution layer and the transmission layer more timely and reduces the overall power supply cost.

[0050] The overall beneficial effects are:

[0051] 1. Based on data collected at local sites, this patent analyzes and constructs a predictive mathematical model for the output characteristics of thermal power generation, distributed wind power generation, distributed photovoltaic power generation, and interruptible loads by considering multiple environmental factors, such as the ramp / slip rate of thermal generators, and combining the Weibull distribution model to fit and estimate the measured average wind speed and standard deviation. This approach addresses the difficulty of dynamically constructing mathematical models.

[0052] 2. Combining the loss costs, power consumption, and transmission characteristics of various distributed power sources, a mathematical model of the transmission and distribution layers is constructed, and the distribution and transmission layers are abstractly separated. At the same time, a dynamic balance loss function for the "transmission-distribution" architecture is proposed, and an accelerated penalty term and convergence iterative algorithm are proposed to make the dynamic response between the distribution and transmission layers more timely, in order to solve the problems of insufficient feature description accuracy and low dynamic convergence speed or high convergence difficulty of existing algorithms. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] Figure 1 This is the flowchart of this patent;

[0054] Figure 2 An example diagram of an optimization example solved by applying the model described in this patent;

[0055] Figure 3 This is a comparison chart of iterative convergence analysis between the model proposed in this patent and the traditional model algorithm.

[0056] Figure 4 It is a real-time control interface for the actual application of the model after training convergence. DETAILED DESCRIPTION

[0057] The present invention is further described in detail below with reference to the accompanying drawings and specific embodiments:

[0058] The present invention proposes a method for rapid coordinated dispatching of distribution networks taking into account the participation of distributed resources on the demand side, aiming to solve the problem of difficulty in dynamically constructing mathematical models and the problems of insufficient feature description accuracy and low dynamic convergence speed or difficulty in convergence of existing algorithms. Figure 1 It is a flow chart of the present invention. As can be seen from the figure, the specific implementation steps of the method are:

[0059] Step 1. Select a model construction area, analyze the output characteristics of conventional thermal power generation, distributed wind power generation, distributed photovoltaic power generation, and interruptible loads in the "transmission-distribution" power grid architecture, and construct a predictive mathematical model;

[0060] Step 2. Combine the loss costs of each power generation form in step 1 to construct a mathematical model of the transmission layer;

[0061] Step 3. Build a corresponding mathematical model based on the power consumption and transmission characteristics of the transmission and distribution layers.

[0062] Step 4. Construct a dynamic balance loss function for the transmission-distribution grid architecture, propose an acceleration penalty term and convergence iterative algorithm, and continuously train the model until convergence, making the dynamic response between the distribution layer and the transmission layer more timely.

[0063] Step 5. Apply the trained model to the distribution network containing distributed resources.

[0064] The specific steps for constructing the mathematical model for distributed power generation prediction in step 1 are as follows:

[0065] Step 1.1. A conventional thermal power generation model is constructed based on the ramp rate and ramp rate of the corresponding generator set. The mathematical model is:

[0066]

[0067] in, and represents the upper / lower limit of the output power of the i-th generator set at time t; and They represent the maximum output power, the power at time t, and the minimum output power of the i-th generator set respectively; DR represents the maximum limit range of power that the i-th unit can deliver at time t. i and UR i Respectively represent the maximum ramp rate and ramp rate of the i-th unit. Through the construction of the mathematical model, the transmission capacity of the generator set at time t is finally determined, thus realizing the adaptive adjustment of the model.

[0068] Step 1.2. Construct a mathematical model for wind speed probability density prediction. This patent adopts a two-parameter Weibull distribution model and introduces the measured value of the wind field average wind speed and the fitting estimate of the standard deviation to determine the model expression:

[0069]

[0070] Where, and They represent the power provided by the i-th wind turbine at time t and the rated output power, V and V respectively. R are wind speed sequence set and rated wind speed, V ci,t and V co,t are the cut-in / cut-out wind speeds at time t respectively.

[0071] Step 1.3. Construct a mathematical model for distributed photovoltaic power generation. This patent uses the maximum likelihood estimation algorithm for the light intensity Beta to determine the photovoltaic timing characteristics, where the photovoltaic output probability density function expression is:

[0072]

[0073] Where, and They represent the maximum illumination intensity per unit area of ​​the i-th wind turbine and the illumination intensity at time t, respectively. α and β represent shape coefficients, and Γ() represents the Gamma function.

[0074] Combined with the influence of the light-receiving area of ​​the photovoltaic array, the output power of the photovoltaic array is finally determined The mathematical expression is as follows:

[0075]

[0076] Where A represents the photovoltaic array area, η c,t Represents the photovoltaic conversion efficiency of the photovoltaic array at time t, K c is the threshold function.

[0077] Step 1.4. Construct a mathematical model for interruptible load power generation. For interruptible loads, the regulation range is the widest, and the maximum discharge power of the i-th device is recorded as Then the discharge power of the i-th device at time t is The range is

[0078] The specific description of the mathematical model of the transmission layer in step 2 is:

[0079] Construct a mathematical model of the transmission layer to minimize the total operating cost and visualize the objective function C dis The mathematical expression is:

[0080] minC dis =C pun +C the +C sun +C the,b +C us -C sell

[0081] Where C pun and C us C represents the penalty cost when the power of the distribution network tie line exceeds the limit and the risk of wind curtailment occurs, the Represents the power generation cost of thermal generators, C sun represents the loss cost of distributed photovoltaic power generation, C the,b represents the backup cost of thermal generators, C sellFor electricity sales revenue.

[0082] The specific expressions of each term in the above formula are as follows:

[0083]

[0084] T represents the total operation time of the distribution layer, N J and N K They represent the thermal power generator set and the distribution network set respectively; ρ pun Indicates the penalty coefficient for exceeding the limit of transmission power of tie line, represents the transmission power of the tie line of the jth thermal generator at time t; m j 、h j and s j They represent the cost coefficient of the j-th thermal generator, is the power generation of the jth thermal generator at time t, ξ sun represents the loss coefficient of distributed photovoltaic power generation, and Represent the positive / negative spinning reserve coefficient, and denote the positive / negative spinning reserve cost coefficient, ρ us 、 and They represent the wind power curtailment penalty coefficient, the dispatch output at time t, and the predicted output at time t, respectively. and They are respectively represented as the node electricity price and boundary transmission power of the kth distribution network at time t.

[0085] The specific description of step 3, which constructs the mathematical model of the distribution layer, is:

[0086] Construct a mathematical model of the distribution layer, and the cost of using the kth distribution layer Mainly from the following aspects: the cost of power transmission in the distribution network Energy storage charging and discharging costs Interruptible charge control cost Power generation cost of generator sets Therefore, in order to improve economic efficiency, it is necessary to minimize The objective function of optimization is:

[0087]

[0088] The specific mathematical expression is as follows:

[0089]

[0090] in represents the power received by the distribution layer at time t, and Represent the discharge / charge cost coefficient, and Respectively represent the energy storage discharge / charging power, ρ cntl and denote the control cost coefficient of the oth interruptible load and the attenuation power at time t, respectively. and They represent the cost coefficient of the jth thermal generator in the kth distribution layer, represents the generator power at time t.

[0091] Among them, the specific steps of step 4: constructing the dynamic balance loss function of the "transmission-distribution" power grid architecture and training and converging the model are:

[0092] Step 4.1. For the kth distribution network, at time t, the ideal boundary transmission power is Equal to the power received by the distribution layer These two variables are in dynamic adjustment. In order to make the loss function capture and accelerate convergence faster, this patent and Based on the difference, a new e-exponential acceleration penalty term is proposed to improve the response speed of the distribution network. The specific objective function can be expressed as:

[0093]

[0094] Where y k,t and z k,t They represent the difference term penalty factor and the acceleration term penalty factor respectively. In this patent, the initial values ​​of iteration are 0.01 and 0.011, and the superscript “—” represents a known term.

[0095] Step 4.2. Similarly, to ensure that the transmission network can quickly obtain input from the distribution network, the transmission network needs to be able to quickly feedback the current distribution network transmission gap. The same principle as step 4.1 is used to construct the corresponding objective function for the transmission network in a collective manner as follows:

[0096]

[0097] Where N q Represents a collection of transmission networks.

[0098] Step 4.3. Formulate the convergence criteria for model iteration. From the perspective of the consistency of dynamic adjustment of distribution network and transmission network, this patent proposes a new convergence algorithm. The specific expression is as follows:

[0099]

[0100] Where, represents the objective function value corresponding to the area, ε1 and ε2 represent the iterative convergence coefficients, which are respectively set as 1-e -2 and e -2 , Item can guarantee and In the case of large differences, the objective function can quickly obtain the difference and converge. Energy Guarantee and In the case of small difference, Supplement it to ensure that the objective function can converge quickly at different stages.

[0101] Step 4.4. For those that do not meet the convergence conditions in step 4.3, update the penalty factor, and then loop through steps 4.3 and 4.4 until the model converges. The update criteria are as follows:

[0102]

[0103] Wherein, ψ and χ represent constants, and their value ranges are [1.8, 2.1] and [1, 3] respectively. They are set as 1.9 and 2.0 in this patent.

[0104] Figure 2 This diagram illustrates an optimization example for the model described in this patent. As can be seen from the diagram, the example involves a transmission network and four distribution networks. The transmission network's power comes from thermal power generators and distributed power sources. Nodes 1, 2, 11, and 16 are connected to distributed power sources, while nodes 5 and 20 are connected to thermal power generators. Furthermore, nodes 1, 23, 24, and 29 connect to distribution networks 1, 2, 3, and 4, respectively. For distribution network 1, an interruptible load is connected at node 15, and a distributed power source at node 29. For distribution network 2, distributed power sources are connected at nodes 2, 8, 9, 13, and 24, an interruptible load is connected at node 18, and a distributed photovoltaic generator is connected at node 22. For distribution network 3, distributed power sources are connected at node 8, a wind turbine is connected at node 13, and an interruptible load is connected at node 27. For distribution network 4, distributed power sources are connected at nodes 9, 16, 20, and 28, a distributed wind turbine is connected at node 7, an interruptible load is connected at node 12, and a distributed photovoltaic generator is connected at node 31. In the entire calculation example, the strategy for real-time electricity prices adopts the time-of-use electricity price mechanism, and the compensation price for interruptible loads in the distribution network and transmission network is positioned at 1.08 times the real-time electricity price. The entire algorithm optimization solution is based on Matlab R2023b, and the convergence training of the model is performed by calling the solver Cplex for calculation. Figure 2It can be seen that a transmission network interacts with multiple distribution networks. In the entire power system, the architecture layering of the distribution network and the transmission network is completed. Then, by constructing the loss function between them, the power distribution between the distribution network and the transmission network is quickly and dynamically adjusted to improve the response speed and reduce the power dispatching cost.

[0105] Figure 3 This is a comparison chart of the iterative convergence analysis between the model proposed in this patent and the traditional model algorithm. Figure 2 The example shown in the figure is used for benchmark analysis. Under the condition of meeting power demand scheduling, the algorithm is solved using the traditional model, the architecture layered model proposed in this patent, and the architecture layered + acceleration penalty factor model. The trend of iteration cost change is shown in the figure below. Figure 2 As shown in the figure, it can be seen that the traditional model obtains a power supply cost of 13,455 yuan after 21 algorithm iterations; this patent proposes an architecture layered model, which obtains a power supply cost of 13,310 yuan after 14 algorithm iterations; the architecture layered + acceleration penalty factor model proposed in this patent converges after only 9 algorithm iterations, and the final power supply cost is 13,110, which proves the effectiveness of the distribution network-transmission network layered architecture proposed in this patent. At the same time, by adding the acceleration penalty term, the algorithm can converge faster, and because of the rapid convergence, the power supply cost is further reduced.

[0106] Figure 4 It is a real-time control interface for the actual application of the model after training convergence, and the Hebei xx station is selected. As can be seen from the figure, the characteristic curves of photovoltaic, distributed power sources and external power consumption vary greatly over time. The use of the model proposed in this patent to dynamically adjust the distribution layer and the transmission layer can well guarantee the power demand. Specifically: At the current moment, the upper reserve capacity of the entire system is 153443.58 kWh, and the lower reserve capacity is 118328.64 kWh, corresponding to the upper / lower reserve rates of 124.17% and 95.75% respectively. At this moment, the forecast error (the control margin of the distribution network) is -3468.43 kWh, and the control error (the control margin of the transmission network) is 5023.91 kWh. Under the background of high power consumption, through the dynamic adjustment between the distribution network and the transmission network, the power demand is met at the lowest power cost.

Claims

1. A fast coordinated dispatching method for distribution networks considering the participation of distributed resources on the demand side, characterized in that: The method comprises the following steps: Step 1. Select a model construction area, analyze the output characteristics of conventional thermal power generation, distributed wind power generation, distributed photovoltaic power generation, and interruptible loads in the "transmission-distribution" grid architecture, and construct a predictive mathematical model; Step 2: Build a mathematical model for the transmission layer to ensure grid power demand while maximizing economic benefits, i.e., minimizing total operating costs. Step 3. Construct a mathematical model of the distribution layer; Step 4. Construct a dynamic balance loss function for the transmission-distribution grid architecture, so that the transmission and distribution networks can meet the demands of both parties while minimizing their own losses. Model this dynamic process: Step 5. Apply the trained distributed power collaborative scheduling model with a differentiated "transmission-distribution" architecture to the distribution network containing distributed resources to improve the dynamic response speed of distribution network scheduling and the economic benefits it brings.

2. The method for rapid coordinated dispatching of distribution networks considering the participation of demand-side distributed resources according to claim 1, characterized in that: Step 1 is as follows: Step 1.

1. Construct the corresponding generator set based on its ramp rate and ramp rate; Step 1.

2. Construct a mathematical model for wind speed probability density prediction; Step 1.

3. Construct a mathematical model of distributed photovoltaic power generation; Step 1.

4. Construct a mathematical model of interruptible load power generation.

3. The method for rapid coordinated dispatching of distribution networks considering the participation of demand-side distributed resources according to claim 2, characterized in that: The mathematical model constructed in step 1.1 according to the climbing rate and sliding rate of the corresponding generator set is: in, and represents the upper / lower limit of the output power of the i-th generator set at time t; and They represent the maximum output power, the power at time t, and the minimum output power of the i-th generator set respectively; DR represents the maximum limit range of power that the i-th unit can deliver at time t. i and UR i They represent the maximum ramp rate and ramp rate of the i-th unit respectively. By constructing the mathematical model, the transmission capacity of the generator set at time t is finally determined, thereby realizing the adaptive adjustment of the model.

4. The method for rapid coordinated dispatching of distribution networks considering the participation of demand-side distributed resources according to claim 2, characterized in that: The mathematical model for wind speed probability density prediction in step 1.2 is: The two-parameter Weibull distribution model is adopted, and the measured value of the wind field average wind speed and the fitting estimate of the standard deviation are introduced to determine the model expression: Where, and They represent the power provided by the i-th wind turbine at time t and the rated output power, V and V respectively. R are wind speed sequence set and rated wind speed, V ci,t and V co,t are the cut-in / cut-out wind speeds at time t respectively.

5. The method for rapid coordinated dispatching of distribution networks considering the participation of demand-side distributed resources according to claim 2, characterized in that: The step 1.3 constructs a distributed photovoltaic power generation mathematical model as follows: The maximum likelihood estimation algorithm is used to determine the photovoltaic timing characteristics of the light intensity Beta, where the photovoltaic output probability density function expression is: Where, and denote the maximum illumination intensity per unit area of ​​the i-th wind turbine and the illumination intensity at time t, respectively; α and β represent shape coefficients; and Γ() represents the Gamma function; Combined with the influence of the light-receiving area of ​​the photovoltaic array, the output power of the photovoltaic array is finally determined The mathematical expression is as follows: Where A represents the photovoltaic array area, η c,t Represents the photovoltaic conversion efficiency of the photovoltaic array at time t, K c is the threshold function.

6. The method for rapid coordinated dispatching of distribution networks considering the participation of demand-side distributed resources according to claim 2, characterized in that: The step 1.4 of constructing the mathematical model of interruptible load power generation is specifically as follows: For interruptible loads, the regulation range is the widest, and the maximum discharge power of the i-th device is recorded as Then the discharge power of the i-th device at time t is The range is 7. The method for rapid coordinated dispatching of distribution networks considering the participation of demand-side distributed resources according to claim 1, characterized in that: Step 2 Objective function C of the mathematical model dis The mathematical expression is: minC dis =C pun +C the +C sun +C the,b +C us -C sell Where C pun and C us C represents the penalty cost when the power of the distribution network tie line exceeds the limit and the risk of wind curtailment occurs, the Represents the power generation cost of thermal generators, C sun represents the loss cost of distributed photovoltaic power generation, C the,b represents the backup cost of thermal generators, C sell Revenue from electricity sales; The specific expressions of each term in the above formula are as follows: T represents the total operation time of the distribution layer, N J and N K They represent the thermal power generator set and the distribution network set respectively; ρ pun Indicates the penalty coefficient for exceeding the limit of transmission power of tie line, represents the transmission power of the tie line of the jth thermal generator at time t; m j 、h j and s j They represent the cost coefficient of the j-th thermal generator, is the power generation of the jth thermal generator at time t, ξ sun represents the loss coefficient of distributed photovoltaic power generation, and Represent the positive / negative spinning reserve coefficient, and denote the positive / negative spinning reserve cost coefficient, ρ us 、 and They represent the wind power curtailment penalty coefficient, the dispatch output at time t, and the predicted output at time t, respectively. and They are respectively represented as the node electricity price and boundary transmission power of the kth distribution network at time t.

8. The method for rapid coordinated dispatching of distribution networks considering the participation of demand-side distributed resources according to claim 1, characterized in that: Step 3 is as follows: The usage cost of the kth distribution layer Mainly from the following aspects: the cost of power transmission in the distribution network Energy storage charging and discharging costs Interruptible charge control cost Power generation cost of generator sets Therefore, in order to improve economic efficiency, it is necessary to minimize The objective function of optimization is: The specific mathematical expression is as follows: in represents the power received by the distribution layer at time t, and Represent the discharge / charge cost coefficient, and Respectively represent the energy storage discharge / charging power, ρ cntl and denote the control cost coefficient of the oth interruptible load and the attenuation power at time t, respectively. and They represent the cost coefficient of the jth thermal generator in the kth distribution layer, represents the generator power at time t.

9. The method for rapid coordinated dispatching of distribution networks considering the participation of demand-side distributed resources according to claim 1, characterized in that: Step 4 is as follows: Step 4.

1. For the kth distribution network, at time t, the ideal boundary transmission power is Equal to the power received by the distribution layer These two variables are in dynamic adjustment. In order to make the loss function capture and accelerate convergence faster, and Based on the difference, a new e-exponential acceleration penalty term is proposed to improve the response speed of the distribution network. The specific objective function can be expressed as: Where y k,t and z k,t They represent the difference term penalty factor and the acceleration term penalty factor, respectively. The initial values ​​of the iteration are 0.01 and 0.

011. The superscript "—" represents a known term. Step 4.

2. Similarly, to ensure that the transmission network can quickly obtain input from the distribution network, the transmission network needs to be able to quickly feedback the current distribution network transmission gap. The same principle as step 4.1 is used to construct the corresponding objective function for the transmission network in a collective manner as follows: Where N q represents the set of transmission networks; Step 4.

3. Establish convergence criteria for model iteration; Step 4.

4. For those that do not meet the convergence conditions in step 4.3, update the penalty factor, and then loop through steps 4.3 and 4.4 until the model converges. The update criteria are as follows: Where ψ and χ are constants with values ​​ranging from [1.8, 2.1] and [1, 3], respectively, and are set to 1.9 and 2.

0.

10. The method for rapid coordinated dispatching of distribution networks considering the participation of demand-side distributed resources according to claim 1, characterized in that: The convergence criterion for model iteration in step 4.3 is: From the perspective of the consistency of dynamic adjustment of distribution network and transmission network, a new convergence algorithm is proposed. The specific expression is as follows: Where, represents the objective function value corresponding to the area, ε1 and ε2 represent the iterative convergence coefficients, which are set to 1-e respectively. -2 and e -2 , Item can guarantee and In the case of large differences, the objective function can quickly obtain the difference and converge. Energy Guarantee and In the case of small difference, Supplement it to ensure that the objective function can converge quickly at different stages.

Images