Integrated Prediction and Scheduling Method and System for Active Distribution Network Considering Power Auxiliary Services

Abstract

The present invention discloses an integrated prediction and scheduling method and system for an active distribution network considering power auxiliary services. The method includes: collecting meteorological forecast data of the active distribution network, and constructing a net load prediction model for each node in the active distribution network considering distributed photovoltaics; based on the net load prediction model, constructing an active and reactive power intraday scheduling model for the active distribution network considering two types of power auxiliary services, namely demand response active power support and distributed photovoltaic reactive power support; combining the net load prediction model and the active and reactive power intraday scheduling model of the active distribution network to form an integrated prediction and scheduling model; solving the integrated prediction and scheduling model to obtain an integrated prediction and scheduling agent; and outputting an intraday active and reactive power scheduling plan for the active distribution network based on the integrated prediction and scheduling agent. The method of the present invention can simplify the source-load prediction stage of intraday active and reactive power scheduling, provide an intraday active and reactive power scheduling plan more quickly and efficiently, and improve the execution efficiency of intraday scheduling of the active distribution network.

Description

Technical Field

[0001] The present invention relates to an integrated predictive dispatching method and system for an active distribution network considering power auxiliary services, belonging to the technical field of power system operation regulation. Background Technique

[0002] Developing a new modern power system with renewable energy generation is one of the key measures to achieve the current energy strategy. For the new modern power system, the active distribution network can efficiently and locally absorb distributed renewable energy such as distributed photovoltaic, so the active distribution network has also become the core link to achieve the energy goal. However, with the large-scale access of distributed renewable energy, its non-stationary characteristics lead to the need for more flexible power dispatching measures for the active distribution network, including promoting the coordinated dispatching of sources, networks, loads, and storage, and improving power auxiliary services. In this case, the active distribution network has strong non-stationary characteristics on both the power generation and consumption sides, which greatly affects the development of power system operation regulation technology.

[0003] With the development of power system operation regulation technology, how to further shorten the gap between the existing dispatching level and the optimal dispatching level under ideal conditions is the core issue in this technical field. Usually, power dispatching operations follow the method of predicting first and then optimizing dispatching. Therefore, dispatching operators cannot judge the impact of different prediction results on dispatching, and at the same time, power forecasters cannot evaluate the gap between their predictions and the optimal dispatching level. Especially in the context of the current need for more flexible power dispatching measures, this core issue is even more difficult to solve. Summary of the Invention

[0004] Object of the Invention: Aiming at the deficiencies in the active distribution network dispatching in the current power system operation regulation technology field, the present invention proposes an integrated predictive dispatching method for an active distribution network considering power auxiliary services, simplifies the source-load prediction stage of intraday active and reactive power dispatching, provides an intraday active and reactive power dispatching plan more quickly and efficiently, improves the execution efficiency and economic benefits of intraday dispatching of the active distribution network, thereby increasing the new energy consumption capacity of the power system and enhancing the competitiveness of power grid enterprises.

[0005] Technical Solution: To achieve the above object of the invention, the integrated predictive dispatching method for an active distribution network considering power auxiliary services proposed by the present invention includes the following steps:

[0006] Step 1: Collect meteorological forecast data of the active distribution network, and construct a net load prediction model considering distributed photovoltaic for each node in the active distribution network. The formula of this model is:

[0007]

[0008] In the formula, P i,t+1:t+Tis the predicted value of the net load of the \(i\)-th node in the active distribution network from time \(t + 1\) to \(t+T\), \(M\) pred,i is the net load prediction model of the \(i\)-th node in the active distribution network, \(W\) i,t-D:t+T is the meteorological forecast data collected by the active distribution network from time \(t - D\) to \(t+T\) is the load value of the \(i\)-th node in the active distribution network from time \(t - D\) to \(t\) is the distributed photovoltaic power value of the \(i\)-th node in the active distribution network from time \(t - D\) to \(t\) is the predicted value of the load of the \(i\)-th node in the active distribution network from time \(t + 1\) to \(t+T\) is the predicted value of the distributed photovoltaic power of the \(i\)-th node in the active distribution network from time \(t + 1\) to \(t+T\), \(D\) is the acquisition step, and \(T\) is the prediction step

[0009] Step 2: Based on the net load prediction models of each node in the active distribution network, construct an active and reactive intra-day scheduling model for the active distribution network considering two types of power auxiliary services: demand response active power support and distributed photovoltaic reactive power support. The formula of this model is expressed as:

[0010]

[0011] In the formula, is the active scheduling cost of the active distribution network at time \(t + k\), \(k\) is the step index is the active scheduling power of the active distribution network at time \(t + k\) is the reactive scheduling cost of the active distribution network at time \(t + k\) is the reactive scheduling power of the active distribution network at time \(t + k\) is the distributed photovoltaic reactive power support cost at time \(t + k\) is the distributed photovoltaic reactive power support power of the \(i\)-th node in the active distribution network at time \(t + k\) is the demand response active power support cost at time \(t + k\) is the demand response active power support power of the \(i\)-th node in the active distribution network at time \(t + k\), \(\sum\) is the summation function

[0012] Step 3: Combine the net load prediction model and the active and reactive intra-day scheduling model of the active distribution network to form an integrated prediction and scheduling model, and unify the security constraint conditions. The formula of the integrated prediction and scheduling model is expressed as:

[0013]

[0014] s.t. \(C\) i,t+k (s t , a t )\(\leq0\)

[0015] Where, F t+k is the objective function of active power and reactive power scheduling within a day for the active distribution network at time t + k, s t is the operating state of the active distribution network at time t. This operating state is a subset of the state set including the weather forecast data collected by the active distribution network from time t - D to t + T, the load values and distributed photovoltaic power values of the i-th node in the active distribution network from time t - D to t. D is the collection step size, and T is the prediction step size; a t is the scheduling action of the active distribution network at time t. This scheduling action is a subset of the action set including the active power scheduling power, reactive power scheduling power, distributed photovoltaic reactive power support power, and demand response active power support power of the active distribution network at time t + k; C i,t+k is the security constraint of the i-th node in the active distribution network at time t + k;

[0016] Step 4: Solve the integrated prediction and scheduling model using the deep deterministic policy gradient algorithm driven by a multi-environment model to obtain an integrated prediction and scheduling agent. The formula of this agent is expressed as:

[0017]

[0018] Where, Q π is the evaluation model of the integrated prediction and scheduling agent, π is the action strategy of the active distribution network, s is the operating state of the active distribution network, a is the scheduling action of the active distribution network, is the expected value function, γ is the discount factor, π * is the policy model of the integrated prediction and scheduling agent, and argmax is the function for finding the parameter of the maximum value;

[0019] Step 5: Based on the integrated prediction and scheduling agent, output the active power and reactive power scheduling plan within a day for the active distribution network to achieve active power and reactive power scheduling within a day. The formula is expressed as:

[0020]

[0021] Where, A is the active power and reactive power scheduling plan within a day for the active distribution network.

[0022] Among them, the net load prediction model M pred,i is a deep neural network model. The input data includes the weather forecast from t - D to t + T, the load values from time t - D to t, and the distributed photovoltaic power values from time t - D to t. The output data includes the load prediction values from time t + 1 to t + T and the distributed photovoltaic power prediction values from time t + 1 to t + T.

[0023] The constraint conditions of the active power and reactive power scheduling model within a day for the active distribution network are:

[0024]

[0025] Wherein, V i is the voltage amplitude of the i-th node in the active distribution network, V min is the minimum constraint of the voltage amplitude, V max is the maximum constraint of the voltage amplitude, is the energy storage power value of the i-th node in the active distribution network at time t + k, P i bm is the energy storage power constraint of the i-th node in the active distribution network, P i dm is the active power support power constraint of the demand response of the i-th node in the active distribution network, is the distributed photovoltaic power value of the i-th node in the active distribution network at time t + k, is the distributed photovoltaic reactive power value of the i-th node in the active distribution network at time t + k, is the complex power constraint of the distributed photovoltaic of the i-th node in the active distribution network, arctan is the arctangent function, is the distributed photovoltaic phase angle constraint of the i-th node in the active distribution network.

[0026] The unified security constraint condition is:

[0027]

[0028] Wherein, C i,t+k is the security constraint of the i-th node in the active distribution network at time t + k, is the conditional summation operation, indicating the sum of each row of constraint conditions on the right; (·≥·) and (·<·) are conditional judgment operations, indicating that the value is 1 when the inequality condition is satisfied, otherwise the value is 0.

[0029] Preferably, the integrated prediction and scheduling model is solved by using the deep deterministic policy gradient algorithm driven by a multi-environment model, including the following steps:

[0030] S1. Construct M environmental models of the active distribution network, and the formula of this model is expressed as:

[0031] s t+1 = M φ,m (s t , a t ), 1 ≤ m ≤ M

[0032] Wherein, M φ,m is the m-th environmental model of the active distribution network, m is the environmental model index, M is the number of environmental models, s t+1 is the operating state of the active distribution network at time t + 1, s t is the operating state of the active distribution network at time t, at The scheduling action of the active distribution network at time t;

[0033] S2. Based on the historical operation samples of the active distribution network, solve the environment model of the active distribution network;

[0034] S3. Based on the deep deterministic policy gradient algorithm, input the environment models of M active distribution networks into the integrated prediction and scheduling model and solve it.

[0035] Among them, the solution formula for the environment model of the active distribution network in step S2 is:

[0036]

[0037] In the formula, N s is the number of historical operation samples of the active distribution network, and is the sum of squares function.

[0038] The solution formula for the integrated prediction and scheduling model in step S3 is:

[0039]

[0040] s.t.C i,t+k (M φ,m (s t ,a t ))≤0, 1≤m≤M

[0041] In the formula, φ DDPG is the parameter of the deep deterministic policy gradient algorithm.

[0042] The present invention also provides an integrated prediction and scheduling system for an active distribution network considering power auxiliary services, including:

[0043] A net load prediction model construction module, which is used to collect meteorological forecast data of the active distribution network and construct a net load prediction model considering distributed photovoltaics for each node in the active distribution network. The formula of this model is expressed as:

[0044]

[0045] In the formula, P i,t+1:t+T is the predicted value of the net load of the i-th node in the active distribution network from time t + 1 to t + T, M pred,i is the net load prediction model of the i-th node in the active distribution network, W i,t-D:t+T is the meteorological forecast data collected by the active distribution network from time t - D to t + T, is the load value of the i-th node in the active distribution network from time t - D to t, is the distributed photovoltaic power value of the i-th node in the active distribution network from time t - D to t, is the load prediction value of the i-th node in the active distribution network from time t+1 to t+T. is the predicted distributed photovoltaic power value of the i-th node in the active distribution network from time t+1 to t+T, D is the acquisition step, and T is the prediction step;

[0046] The active and reactive intraday scheduling model construction module is used to construct an active and reactive intraday scheduling model for the active distribution network considering two types of power auxiliary services, namely demand response active support and distributed photovoltaic reactive support, based on the net load prediction model of each node in the active distribution network. The formula of this model is expressed as:

[0047]

[0048] In the formula, is the active scheduling cost of the active distribution network at time t+k, k is the step index, is the active scheduling power of the active distribution network at time t+k, is the reactive scheduling cost of the active distribution network at time t+k, is the reactive scheduling power of the active distribution network at time t+k, is the distributed photovoltaic reactive support cost at time t+k, is the distributed photovoltaic reactive support power of the i-th node in the active distribution network at time t+k, is the demand response active support cost at time t+k, is the demand response active support power of the i-th node in the active distribution network at time t+k, Σ is the summation function;

[0049] The integrated prediction and scheduling model construction module is used to combine the net load prediction model and the active and reactive intraday scheduling model of the active distribution network to form an integrated prediction and scheduling model and unify the security constraint conditions. The formula of the integrated prediction and scheduling model is expressed as:

[0050]

[0051] s.t.C i,t+k (s t ,a t )≤0

[0052] In the formula, F t+k is the intraday active and reactive scheduling objective function of the active distribution network at time t+k, s t is the operating state of the active distribution network at time t. This operating state is a subset of the state set including the meteorological forecast data collected by the active distribution network from time t-D to t+T, the load values and distributed photovoltaic power values of the i-th node in the active distribution network from time t-D to t. D is the acquisition step, and T is the prediction step; a tThe scheduling action of the active distribution network at time t, which is a subset of the action set including the active scheduling power, reactive scheduling power, distributed PV reactive support power, and demand response active support power of the active distribution network at time t + k; C i,t+k The security constraint of the i-th node in the active distribution network at time t + k;

[0053] The integrated prediction and scheduling model solving module is used to solve the integrated prediction and scheduling model by using the deep deterministic policy gradient algorithm driven by a multi-environment model to obtain an integrated prediction and scheduling agent, which is represented by the formula:

[0054]

[0055] In the formula, Q π is the evaluation model of the integrated prediction and scheduling agent, π is the action strategy of the active distribution network, s is the operating state of the active distribution network, a is the scheduling action of the active distribution network, is the expected value function, γ is the discount factor, π * is the policy model of the integrated prediction and scheduling agent, and argmax is the maximum value parameter function;

[0056] The active and reactive power scheduling scheme determination module is used to output the active and reactive power scheduling scheme of the active distribution network within a day based on the integrated prediction and scheduling agent to realize the active and reactive power scheduling within a day, which is represented by the formula:

[0057]

[0058] In the formula, A is the active and reactive power scheduling scheme of the active distribution network within a day.

[0059] The present invention also provides a computer device, including: one or more processors; a memory; and one or more programs, where the one or more programs are stored in the memory and are configured to be executed by the one or more processors, and when the programs are executed by the processors, the steps of the integrated prediction and scheduling method for the active distribution network considering power auxiliary services as described above are implemented.

[0060] The present invention also provides a computer-readable storage medium, on which a computer program is stored, and when the computer program is executed by a processor, the steps of the integrated prediction and scheduling method for the active distribution network considering power auxiliary services as described above are implemented.

[0061] Beneficial effects:

[0062] (1) By collecting the meteorological forecast data of the active distribution network, a net load prediction model considering distributed photovoltaics is constructed for each node in the active distribution network; based on the net load prediction model, an active and reactive power intraday scheduling model for the active distribution network considering two types of power auxiliary services, namely demand response active power support and distributed photovoltaic reactive power support, is constructed; the net load prediction model and the active and reactive power intraday scheduling model of the active distribution network are combined to form an integrated prediction and scheduling model; the integrated prediction and scheduling model is solved to obtain an integrated prediction and scheduling agent; based on the integrated prediction and scheduling agent, an intraday active and reactive power scheduling plan for the active distribution network is output. Compared with the existing active distribution network scheduling technology, the method of the present invention can simplify the source-load prediction stage of intraday active and reactive power scheduling and provide an intraday active and reactive power scheduling plan more quickly and efficiently.

[0063] (2) The method of the present invention can meet the scheduling requirements of the active distribution network including power auxiliary services and ensure the flexibility and stability of the regulation of various resources in the active distribution network.

[0064] (3) Since it can provide an intraday active and reactive power scheduling plan more quickly and efficiently, the method of the present invention helps to improve the execution efficiency and economic benefits of intraday scheduling in the active distribution network, thereby increasing the new energy consumption capacity of the power system and enhancing the competitiveness of power grid enterprises. BRIEF DESCRIPTION OF THE DRAWINGS

[0065] Figure 1 It is a schematic diagram of the principle of the method of the present invention.

[0066] Figure 2 It is a diagram of the predicted distributed photovoltaic power of the method of the present invention.

[0067] Figure 3 It is a diagram of the analysis result of the scheduling cost-benefit of the method of the present invention. DETAILED DESCRIPTION OF THE INVENTION

[0068] The technical solution of the present invention will be further described below with reference to the drawings.

[0069] The present invention proposes an integrated prediction and scheduling method for an active distribution network considering power auxiliary services. As Figure 1 shown, the method includes the following steps:

[0070] (1) Collect the meteorological forecast data of the active distribution network, and construct a net load prediction model considering distributed photovoltaics for each node in the active distribution network;

[0071] (2) Based on the net load prediction model of each node in the active distribution network, construct an active and reactive power intraday scheduling model for the active distribution network considering two types of power auxiliary services, namely demand response active power support and distributed photovoltaic reactive power support;

[0072] (3) Combine the net load prediction model and the active distribution network active and reactive power intraday scheduling model to form an integrated prediction and scheduling model, and unify the security constraint conditions;

[0073] (4) Use the deep deterministic policy gradient algorithm driven by a multi-environment model to solve the integrated prediction and scheduling model to obtain an integrated prediction and scheduling agent;

[0074] (5) Based on the integrated prediction and scheduling agent, output the active and reactive power intraday scheduling plan for the active distribution network to achieve intraday active and reactive power scheduling.

[0075] The following combines specific implementation cases to detail the specific implementation process of using the method in the present invention for intraday active and reactive power scheduling of an active distribution network. The present invention selects the standard distribution network example of IEEE-33 nodes, connects a distributed photovoltaic and energy storage system to each load node in the example, the acquisition step size D = 72 hours, and the prediction step size T = 24 hours.

[0076] Based on the above case, the specific implementation steps of the method of the present invention are as follows:

[0077] Step 1: Collect the meteorological forecast data of the active distribution network, and construct a net load prediction model considering distributed photovoltaics for each node in the active distribution network. The formula of this model is expressed as:

[0078]

[0079] In the formula, P i,t+1:t+T is the predicted value of the net load of the i-th node in the active distribution network from time t + 1 to t + T, M pred,i is the net load prediction model of the i-th node in the active distribution network, W i,t-D:t+T is the meteorological forecast data collected by the active distribution network from time t - D to t + T, is the load value of the i-th node in the active distribution network from time t - D to t, is the distributed photovoltaic power value of the i-th node in the active distribution network from time t - D to t, is the predicted load value of the i-th node in the active distribution network from time t + 1 to t + T, is the predicted distributed photovoltaic power value of the i-th node in the active distribution network from time t + 1 to t + T, D is the acquisition step size, and T is the prediction step size. The net load prediction model M pred,i is a deep neural network model. The input data includes the meteorological forecast from t - D to t + T, the load value from time t - D to t, and the distributed photovoltaic power value from time t - D to t. The output data includes the predicted load value from time t + 1 to t + T and the predicted distributed photovoltaic power value from time t + 1 to t + T, as Figure 2The figure shows the distributed photovoltaic power prediction result diagram in the embodiment of the present invention.

[0080] Step 2: Based on the net load prediction models of each node in the active distribution network, construct an active and reactive intra-day scheduling model for the active distribution network considering two types of power auxiliary services, namely demand response active power support and distributed photovoltaic reactive power support. The formula of this model is expressed as:

[0081]

[0082] In the formula, is the active scheduling cost of the active distribution network at time t + k, where k is the step index, is the active scheduling power of the active distribution network at time t + k, is the reactive scheduling cost of the active distribution network at time t + k, is the reactive scheduling power of the active distribution network at time t + k, is the distributed photovoltaic reactive power support cost at time t + k, is the distributed photovoltaic reactive power support power of the i-th node in the active distribution network at time t + k, is the demand response active power support cost at time t + k, is the demand response active power support power of the i-th node in the active distribution network at time t + k, V i is the voltage amplitude of the i-th node in the active distribution network, V min is the minimum constraint of the voltage amplitude, V max is the maximum constraint of the voltage amplitude, is the energy storage power value of the i-th node in the active distribution network at time t + k, P i bm is the energy storage power constraint of the i-th node in the active distribution network, P i dm is the demand response active power support power constraint of the i-th node in the active distribution network, is the distributed photovoltaic power value of the i-th node in the active distribution network at time t + k, is the distributed photovoltaic reactive power value of the i-th node in the active distribution network at time t + k, is the distributed photovoltaic complex power constraint of the i-th node in the active distribution network, arctan is the arctangent function, is the distributed photovoltaic phase angle constraint of the i-th node in the active distribution network, Σ is the summation function;

[0083] Step 3: Use the prediction error in the net load prediction model to calculate the energy storage power in the active and reactive intra-day scheduling model of the active distribution network, so as to realize the combination of the two models. The calculation formula is:

[0084]

[0085] In the formula, is the energy storage power value of the i-th node in the active distribution network at time t + k, and P i,t+k is the predicted value of the net load of the i-th node in the active distribution network at time t + k. is the actual value of the net load of the i-th node in the active distribution network at time t + k. [k] is the output index of the net load prediction model, that is, when predicting the k-th time step, the k-th output of the prediction model is taken;

[0086] Taking the active power dispatch power, reactive power dispatch power, distributed photovoltaic reactive power support power, and demand response active power support power of the active distribution network at time t + k as the action set, and taking the active distribution network states such as weather forecast data, load values, and distributed photovoltaic power values as the state set, an integrated prediction and dispatch model is constructed, and the formula is expressed as:

[0087]

[0088] In the formula, F t+k is the integrated prediction and dispatch objective function of the active distribution network at time t + k; is the subset operator; the unified security constraint conditions are expressed by the formula:

[0089]

[0090] In the formula, C i,t+k is the security constraint of the i-th node in the active distribution network at time t + k; is the conditional summation operation, indicating the sum of each row of constraint conditions on the right side; (·≥·) and (·<·) are conditional judgment operations, indicating that the value is 1 when the inequality condition is satisfied, otherwise the value is 0; then the formula of the integrated prediction and dispatch model is expressed as:

[0091]

[0092] s.t. C i,t+k (s t , a t ) ≤ 0

[0093] In the formula, F t+k is the intraday active and reactive power dispatch objective function of the active distribution network at time t + k, s t is the operating state of the active distribution network at time t, a t is the dispatch action of the active distribution network at time t, C i,t+k is the security constraint of the i-th node in the active distribution network at time t + k, and this constraint value must be less than or equal to 0;

[0094] Step 4: Solve the integrated prediction and scheduling model using the multi-environment model-driven deep deterministic policy gradient algorithm. This algorithm constructs M environment models for active distribution networks, and the model formula is expressed as:

[0095] s t+1 = M φ,m (s t , a t ), 1 ≤ m ≤ M

[0096] In the formula, M φ,m is the m-th environment model of the active distribution network, m is the environment model index, M is the number of environment models, s t+1 is the operating state of the active distribution network at time t + 1, s t is the operating state of the active distribution network at time t, a t is the scheduling action of the active distribution network at time t; Based on the historical operation samples of the active distribution network, solve the environment model of the active distribution network, and the solution formula is:

[0097]

[0098] In the formula, N s is the number of historical operation samples of the active distribution network, and is the sum of squares function; Based on the deep deterministic policy gradient algorithm, substitute the M environment models of the active distribution network into the integrated prediction and scheduling model and solve it. The solution formula is:

[0099]

[0100] s.t. C i,t+k (M φ,m (s t , a t )) ≤ 0, 1 ≤ m ≤ M

[0101] In the formula, φ DDPG is the parameter of the deep deterministic policy gradient algorithm; Obtain the integrated prediction and scheduling agent, and the agent formula is expressed as:

[0102]

[0103] In the formula, Q π is the evaluation model of the integrated prediction and scheduling agent, π is the action strategy of the active distribution network, s is the operating state of the active distribution network, a is the scheduling action of the active distribution network, is the expected value function, γ is the discount factor, π * is the policy model of the integrated prediction and scheduling agent, and argmax is the maximum value parameter function;

[0104] Step 5: Based on the integrated prediction and scheduling agent, output the active and reactive power scheduling plan for the active distribution network within a day to achieve the active and reactive power scheduling within a day, which is expressed by the formula:

[0105]

[0106] In the formula, A is the active and reactive power scheduling plan for the active distribution network within a day.

[0107] The comparison of the calculation time of the integrated prediction and scheduling of the present invention with the traditional prediction and scheduling is shown in Table 1. It can be seen that the present invention can simplify the source-load prediction stage of the active and reactive power scheduling within a day, reduce the total calculation time, and thus improve the execution efficiency.

[0108] Table 1 Comparison results of the calculation time of the prediction and scheduling

[0109]

[0110] In addition, the present invention can improve the economic benefits of the active distribution network within-day scheduling. As Figure 3 shown in the figure of the scheduling cost-benefit analysis result in this embodiment, it can be seen that the method of the present invention can improve the economic benefits of the existing active distribution network within-day scheduling.

[0111] In summary, the integrated prediction and scheduling method for the active distribution network considering power auxiliary services of the present invention can simplify the source-load prediction stage of the active and reactive power scheduling within a day, provide the active and reactive power scheduling plan within a day more quickly and efficiently, meet the scheduling requirements of the active distribution network including power auxiliary services, ensure the flexibility and stability of the regulation of various resources in the active distribution network, and because it can provide the active and reactive power scheduling plan within a day more quickly and efficiently, the method of the present invention helps to improve the execution efficiency and economic benefits of the active distribution network within-day scheduling, thereby increasing the new energy consumption capacity of the power system and enhancing the competitiveness of power grid enterprises.

[0112] Based on the same technical concept as the method embodiment, the present invention also provides an integrated prediction and scheduling system for the active distribution network considering power auxiliary services, including:

[0113] A net load prediction model construction module, which is used to collect the meteorological forecast data of the active distribution network and construct a net load prediction model considering distributed photovoltaic for each node in the active distribution network. The formula of this model is expressed as:

[0114]

[0115] In the formula, P i,t+1:t+T is the predicted value of the net load of the i-th node in the active distribution network from time t + 1 to t + T, M pred,i is the net load prediction model of the i-th node in the active distribution network, W i,t-D:t+Tis the meteorological forecast data collected by the active distribution network from time t - D to t + T, is the load value of the i-th node in the active distribution network from time t - D to t, is the distributed photovoltaic power value of the i-th node in the active distribution network from time t - D to t, is the load prediction value of the i-th node in the active distribution network from time t + 1 to t + T, is the distributed photovoltaic power prediction value of the i-th node in the active distribution network from time t + 1 to t + T, D is the acquisition step, and T is the prediction step;

[0116] The active and reactive intra-day scheduling model construction module is used to construct an active and reactive intra-day scheduling model for the active distribution network considering two types of power auxiliary services, namely demand response active support and distributed photovoltaic reactive support, based on the net load prediction model of each node in the active distribution network. The formula of this model is expressed as:

[0117]

[0118] In the formula, is the active scheduling cost of the active distribution network at time t + k, k is the step index, is the active scheduling power of the active distribution network at time t + k, is the reactive scheduling cost of the active distribution network at time t + k, is the reactive scheduling power of the active distribution network at time t + k, is the distributed photovoltaic reactive support cost at time t + k, is the distributed photovoltaic reactive support power of the i-th node in the active distribution network at time t + k, is the demand response active support cost at time t + k, is the demand response active support power of the i-th node in the active distribution network at time t + k, Σ is the summation function;

[0119] The integrated prediction and scheduling model construction module is used to combine the net load prediction model and the active and reactive intra-day scheduling model of the active distribution network to form an integrated prediction and scheduling model, and unify the security constraint conditions. The formula of the integrated prediction and scheduling model is expressed as:

[0120]

[0121] s.t.C i,t+k (s t ,a t )≤0

[0122] In the formula, F t+k is the intra-day active and reactive scheduling objective function of the active distribution network at time t + k, st is the operating state of the active distribution network at time t, which is a subset of the state set including the weather forecast data collected by the active distribution network from time t - D to t + T, the load values and distributed photovoltaic power values of the i-th node in the active distribution network from time t - D to t, where D is the collection step and T is the prediction step; a t is the scheduling action of the active distribution network at time t, which is a subset of the action set including the active scheduling power, reactive scheduling power, distributed photovoltaic reactive support power, and demand response active support power of the active distribution network at time t + k; C i,t+k is the security constraint of the i-th node in the active distribution network at time t + k;

[0123] The integrated prediction and scheduling model solving module is used to solve the integrated prediction and scheduling model by using the deep deterministic policy gradient algorithm driven by a multi-environment model to obtain an integrated prediction and scheduling agent, which is represented by the following formula:

[0124]

[0125] In the formula, Q π is the evaluation model of the integrated prediction and scheduling agent, π is the action strategy of the active distribution network, s is the operating state of the active distribution network, a is the scheduling action of the active distribution network, is the expected value function, γ is the discount factor, π * is the policy model of the integrated prediction and scheduling agent, and argmax is the maximum value parameter function;

[0126] The active and reactive power scheduling scheme determination module is used to output the active and reactive power scheduling scheme of the active distribution network within a day based on the integrated prediction and scheduling agent to realize the active and reactive power scheduling within a day, which is represented by the following formula:

[0127]

[0128] In the formula, A is the active and reactive power scheduling scheme of the active distribution network within a day.

[0129] It should be understood that the integrated prediction and scheduling system of the active distribution network considering power auxiliary services in the embodiments of the present invention can implement all the technical solutions in the above method embodiments. The functions of its respective functional modules can be specifically implemented according to the methods in the above method embodiments, and the specific implementation process can refer to the relevant descriptions in the above embodiments, which will not be elaborated here.

[0130] The present invention also provides a computer device, including: one or more processors; a memory; and one or more programs, wherein the one or more programs are stored in the memory and configured to be executed by the one or more processors, and when the programs are executed by the processors, the steps of the integrated prediction and scheduling method for an active distribution network considering power auxiliary services as described above are implemented.

[0131] The present invention also provides a computer-readable storage medium, on which a computer program is stored, and when the computer program is executed by a processor, the steps of the integrated prediction and scheduling method for an active distribution network considering power auxiliary services as described above are implemented.

[0132] Those skilled in the art should understand that the embodiments of the present invention can be provided as a method, a device (system), a computer device, or a computer program product. Therefore, the present invention can adopt the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present invention can adopt the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk memories, CD-ROMs, optical memories, etc.) containing computer-usable program codes.

[0133] The present invention is described with reference to the flowchart of the method according to the embodiments of the present invention. It should be understood that each process in the flowchart and the combination of the processes in the flowchart can be implemented by computer program instructions. These computer program instructions can be provided to the processors of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing devices to generate a machine, so that the instructions executed by the processors of the computer or other programmable data processing devices generate a device for implementing the functions specified in one Figure 1 process or multiple processes.

[0134] These computer program instructions can also be stored in a computer-readable memory capable of guiding a computer or other programmable data processing devices to work in a specific manner, so that the instructions stored in the computer-readable memory generate a manufactured product including an instruction device, and the instruction device implements the functions specified in one Figure 1 process or multiple processes.

[0135] These computer program instructions can also be loaded onto a computer or other programmable data processing devices, so that a series of operation steps are executed on the computer or other programmable devices to generate a computer-implemented process, and thus the instructions executed on the computer or other programmable devices provide steps for implementing the functions specified in one Figure 1 process or multiple processes.

Claims

1. An integrated forecasting and dispatching method for active distribution network taking into account power auxiliary services, characterized in that: The following steps are involved: Step 1: Collect the meteorological forecast data of the active distribution network and build a net load prediction model considering distributed photovoltaics for each node in the active distribution network. The formula of the model is expressed as: Where P i,t+1:t+T is the net load forecast value of the ith node in the active distribution network from time t+1 to t+T, M pred,i is the net load prediction model of the ith node in the active distribution network, W i,t-D:t+T is the weather forecast data collected by the active distribution network from time tD to t+T, is the load value of the ith node in the active distribution network from time tD to t, is the distributed photovoltaic power value of the i-th node in the active distribution network from time tD to t, is the load forecast value of the i-th node in the active distribution network from time t+1 to t+T, is the predicted value of distributed photovoltaic power of the i-th node in the active distribution network from time t+1 to t+T, D is the acquisition step length, and T is the prediction step length; Step 2: Based on the net load prediction model of each node in the active distribution network, an active distribution network active and reactive daytime dispatch model is constructed considering two types of power auxiliary services: demand response active support and distributed photovoltaic reactive support. The formula of the model is expressed as follows: In the formula, is the active dispatch cost of the active distribution network at time t+k, k is the step index, is the active dispatching power of the active distribution network at time t+k, is the reactive power dispatch cost of the active distribution network at time t+k, is the reactive dispatching power of the active distribution network at time t+k, is the distributed photovoltaic reactive support cost at time t+k, is the distributed photovoltaic reactive support power of the i-th node in the active distribution network at time t+k, is the active power support cost of demand response at time t+k, is the demand response active support power of the ith node in the active distribution network at time t+k, Σ is the summation function; Step 3: Combine the net load forecasting model and the active distribution network active and reactive daytime dispatching model to form an integrated forecasting and dispatching model, and unify the safety constraints. The formula of the integrated forecasting and dispatching model is expressed as follows: In the formula, F t+k is the active and reactive power dispatch objective function of the active distribution network at time t+k, s t is the operating state of the active distribution network at time t, which is a subset of the state set including the meteorological forecast data collected by the active distribution network from time tD to t+T, the load value of the i-th node in the active distribution network from time tD to t, and the distributed photovoltaic power value, D is the collection step length, and T is the prediction step length; a t is the dispatching action of the active distribution network at time t, which is a subset of the action set including the active dispatching power, reactive dispatching power, distributed photovoltaic reactive support power, and demand response active support power of the active distribution network at time t+k; C i,t+k is the safety constraint of the ith node in the active distribution network at time t+k; Step 4: Use the deep deterministic policy gradient algorithm driven by the multi-environment model to solve the integrated prediction and scheduling model to obtain the integrated prediction and scheduling agent, which is expressed as follows: In the formula, Q π is the evaluation model of the integrated forecasting and dispatching intelligent agent, π is the action strategy of the active distribution network, s is the operating state of the active distribution network, a is the dispatching action of the active distribution network, is the expected value function, γ is the discount factor, π * is the strategy model of the integrated prediction and scheduling agent, and argmax is the maximum value parameter function; Step 5: Based on the integrated prediction and dispatching intelligent body, the active distribution network’s daily active and reactive dispatching scheme is output to achieve daily active and reactive dispatching. The formula is: Where A is the daily active and reactive power dispatching scheme of the active distribution network.

2. The method according to claim 1, characterized in that Net load forecasting model M pred,i It is a deep neural network model. The input data include weather forecast from tD to t+T, load value from time tD to t, and distributed photovoltaic power value from time tD to t. The output data include load prediction value from time t+1 to t+T, and distributed photovoltaic power prediction value from time t+1 to t+T.

3. The method according to claim 1, characterized in that The constraints of the active distribution network active and reactive daytime dispatch model are: Where V i is the voltage amplitude of the ith node in the active distribution network, V min is the minimum constraint on voltage amplitude, V max is the maximum constraint on the voltage amplitude, is the energy storage power value of the i-th node in the active distribution network at time t+k, is the energy storage power constraint of the i-th node in the active distribution network, is the demand response active support power constraint of the ith node in the active distribution network, is the distributed photovoltaic power value of the i-th node in the active distribution network at time t+k, is the distributed photovoltaic reactive power value of the i-th node in the active distribution network at time t+k, is the distributed photovoltaic complex power constraint of the i-th node in the active distribution network, arctan is the inverse tangent function, is the distributed photovoltaic phase angle constraint of the i-th node in the active distribution network.

4. The method according to claim 3, characterized in that The unified safety constraints are: In the formula, C i,t+k is the safety constraint of the ith node in the active distribution network at time t+k, It means to sum up the constraints of each row on the right side; (·≥·) and (·<·) are conditional judgment operations, which means that the value is 1 when the inequality condition is met, and the value is 0 otherwise.

5. The method according to claim 1, characterized in that The integrated prediction and scheduling model is solved using a deep deterministic policy gradient algorithm driven by a multi-environment model. The following steps are involved: S1, construct the environmental model of M active distribution networks, the model formula is expressed as: s t+1 =M φ,m (s t ,a t ),1≤m≤M Where M φ,m is the mth environmental model of the active distribution network, m is the environmental model index, M is the number of environmental models, s t+1 is the operating state of the active distribution network at time t+1, s t is the operating state of the active distribution network at time t, a t is the dispatching action of the active distribution network at time t; S2, based on the historical operation samples of the active distribution network, solve the environmental model of the active distribution network; S3, based on the deep deterministic policy gradient algorithm, brings the environmental model of M active distribution networks into the integrated predictive scheduling model and solves it.

6. The method according to claim 5, characterized in that The solution formula of the environmental model of the active distribution network in step S2 is: Where N s is the number of historical operation samples of the active distribution network, Sum of squares function.

7. The method according to claim 5, characterized in that The solution formula of the integrated forecasting and scheduling model in step S3 is: In the formula, φ DDPG are the parameters of the deep deterministic policy gradient algorithm.

8. An integrated forecasting and dispatching system for active distribution networks taking into account power auxiliary services, characterized in that: include: The net load prediction model building module is used to collect the meteorological forecast data of the active distribution network and build a net load prediction model considering distributed photovoltaics for each node in the active distribution network. The formula of the model is expressed as follows: Where P i,t+1:t+T is the net load forecast value of the ith node in the active distribution network from time t+1 to t+T, M pred,i is the net load prediction model of the ith node in the active distribution network, W i,t-D:t+T is the weather forecast data collected by the active distribution network from time tD to t+T, is the load value of the ith node in the active distribution network from time tD to t, is the distributed photovoltaic power value of the i-th node in the active distribution network from time tD to t, is the load forecast value of the i-th node in the active distribution network from time t+1 to t+T, is the predicted value of distributed photovoltaic power of the i-th node in the active distribution network from time t+1 to t+T, D is the acquisition step length, and T is the prediction step length; The active and reactive daytime dispatch model construction module is used to construct the active and reactive daytime dispatch model of the active distribution network considering two types of power auxiliary services, namely demand response active support and distributed photovoltaic reactive support, based on the net load prediction model of each node in the active distribution network. The formula of the model is expressed as follows: In the formula, is the active dispatch cost of the active distribution network at time t+k, k is the step index, is the active dispatching power of the active distribution network at time t+k, is the reactive power dispatch cost of the active distribution network at time t+k, is the reactive dispatching power of the active distribution network at time t+k, is the distributed photovoltaic reactive support cost at time t+k, is the distributed photovoltaic reactive support power of the i-th node in the active distribution network at time t+k, is the active power support cost of demand response at time t+k, is the demand response active support power of the ith node in the active distribution network at time t+k, Σ is the summation function; The integrated prediction and dispatch model construction module is used to combine the net load prediction model and the active distribution network active and reactive daytime dispatch model to form an integrated prediction and dispatch model and unify the safety constraints. The formula of the integrated prediction and dispatch model is expressed as follows: In the formula, F t+k is the active and reactive power dispatch objective function of the active distribution network at time t+k, s t is the operating state of the active distribution network at time t, which is a subset of the state set including the meteorological forecast data collected by the active distribution network from time tD to t+T, the load value of the i-th node in the active distribution network from time tD to t, and the distributed photovoltaic power value, D is the collection step length, and T is the prediction step length; a t is the dispatching action of the active distribution network at time t, which is a subset of the action set including the active dispatching power, reactive dispatching power, distributed photovoltaic reactive support power, and demand response active support power of the active distribution network at time t+k; C i,t+k is the safety constraint of the ith node in the active distribution network at time t+k; The integrated prediction and scheduling model solving module is used to solve the integrated prediction and scheduling model using a deep deterministic policy gradient algorithm driven by a multi-environment model to obtain an integrated prediction and scheduling agent. The agent formula is expressed as: In the formula, Q π is the evaluation model of the integrated forecasting and dispatching intelligent agent, π is the action strategy of the active distribution network, s is the operating state of the active distribution network, a is the dispatching action of the active distribution network, is the expected value function, γ is the discount factor, π * is the strategy model of the integrated prediction and scheduling agent, and argmax is the maximum value parameter function; The active and reactive power dispatching scheme determination module is used to output the active and reactive power dispatching scheme of the active distribution network within a day based on the integrated prediction and dispatching intelligent body, so as to realize the active and reactive power dispatch within a day. The formula is expressed as follows: Where A is the daily active and reactive power dispatching scheme of the active distribution network.

9. A computer device, characterized in that: include: one or more processors; Memory; And one or more programs, wherein the one or more programs are stored in the memory and are configured to be executed by the one or more processors, and when the programs are executed by the processors, the steps of the active distribution network integrated predictive scheduling method taking into account power auxiliary services as described in any one of claims 1-7 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the active distribution network integrated prediction and dispatching method taking into account power auxiliary services are implemented as described in any one of claims 1 to 7.

Images