CN-122001019-A - Intelligent scheduling decision method based on intelligent auxiliary distribution network scheduling model

CN122001019ACN 122001019 ACN122001019 ACN 122001019ACN-122001019-A

Abstract

The invention discloses an intelligent scheduling decision method based on an intelligent auxiliary model of distribution network scheduling, which relates to the technical field of distribution network scheduling and constructs an intelligent auxiliary model associated with environment by acquiring data such as photovoltaic power generation power, illumination intensity, temperature and the like of a target photovoltaic array from a cloud database; the method comprises the steps of obtaining illumination intensity and temperature by using a meteorological system in real time, predicting a photovoltaic power generation power sequence through a model, performing optimization matching with a predicted load sequence of a power distribution network to generate a scheduling decision instruction set, determining issuing operation of the instruction set according to a regulation and control instruction of an operator, generating a feedback sample set by combining actual illumination intensity, temperature and photovoltaic power generation power, and performing feedback optimization on the model, so that intelligent scheduling decision of the power distribution network is realized, scheduling efficiency and accuracy are improved, and stable operation of the power distribution network is ensured.

Inventors

TANG XINJIE
WANG XIHUA
XU YINGFENG
DAI XIANJUN
CHEN SHIWEN
PAN LI
LIAO BING
DAI FEI
HU HAO
WU SHA
GUO FEI
HE LINJIA

Assignees

安徽知途信息技术有限公司

Dates

Publication Date: 20260508
Application Date: 20260127

Claims (9)

1. The intelligent scheduling decision method based on the intelligent auxiliary distribution network scheduling model is characterized by comprising the following steps of: Step one, obtaining photovoltaic power generation power related to a target photovoltaic array and a time sequence from a cloud database; synchronously extracting illumination intensity and temperature, and constructing a distribution network scheduling intelligent auxiliary model related to the environment of the target photovoltaic array; Acquiring the illumination intensity and the temperature associated with the area where the target photovoltaic array is located in a preset period in real time based on a meteorological system, and predicting the photovoltaic power generation power associated with the target photovoltaic array in the preset period in real time based on an intelligent auxiliary model for distribution network scheduling to generate a predicted photovoltaic power generation power sequence; optimizing and matching based on the predicted photovoltaic power generation power sequence and a predicted load sequence associated with the power distribution network, and generating a power distribution network scheduling decision instruction set; And step four, determining issuing operation of a power distribution network scheduling decision instruction set based on an operator regulation instruction, extracting actual illumination intensity and actual temperature associated with time sequence in a preset period, generating a feedback sample set together with actual photovoltaic power generation power, and performing feedback optimization on an intelligent auxiliary model of power distribution network scheduling.
2. The method according to claim 1, wherein in the first step, the specific manner of obtaining the photovoltaic power generation power associated with the target photovoltaic array and the time sequence from the cloud database is as follows: extracting photovoltaic power generation power of a target photovoltaic array in a tracing period taking the current time as the end time based on a pre-constructed cloud database, wherein the tracing period is a time period preset by an operator; counting the total time in the tracing period, and recording as j; And (3) sequencing the photovoltaic power generation power corresponding to the j moments according to the moments, and recording the photovoltaic power generation power as a historical photovoltaic power generation power sequence GP1, GP 2.
3. The method according to claim 2, wherein in the first step, the specific way of synchronously extracting the illumination intensity and the temperature and constructing the intelligent auxiliary model for the distribution network scheduling of the association of the target photovoltaic array and the environment is as follows: S31, extracting temperature and illumination intensity associated with a time sequence of a region where a target photovoltaic array is located in a tracing period from a cloud database, and processing according to a historical photovoltaic power generation power construction mode to construct a historical temperature sequence T1, T2, tj and a historical illumination intensity sequence L1, L2, lj; S32, extracting a preset sliding influence evaluation window Ht, wherein the duration of the sliding influence evaluation window Ht is m times, m < j, j divides m completely, and j/m=n; s33, performing synchronous division on the historical temperature sequences T1, T2, tj, the historical illumination intensity sequences L1, L2, lj and the historical photovoltaic power generation powers GP1, GP2, GPj based on the sliding influence evaluation window Ht to obtain n historical temperature subsequences, and forming a historical temperature subsequence set according to a time sequence; similarly, a historical illumination intensity subsequence set and a historical photovoltaic power generation power subsequence set are obtained; S34, extracting any one historical temperature subsequence in the historical temperature subsequence set and a corresponding historical illumination intensity subsequence and a historical photovoltaic power generation power subsequence; Calculating a group of dynamic influence factor groups based on the historical temperature and the historical illumination intensity at each moment relative to the historical photovoltaic power generation power at the corresponding moment in the historical photovoltaic power generation power subsequence; s35, determining n dynamic influence factor sets based on the method in the step S34, and constructing the n dynamic influence factor sets into a high-dimensional environment state vector according to time sequence; s36, taking the high-dimensional environment state vector as an input sample, and taking the historical photovoltaic power generation power at the corresponding moment as a prediction target label to form a training sample set; and S37, performing supervised training on the pre-constructed model based on the training sample set until convergence, and recording the model as a distribution network scheduling intelligent auxiliary model, wherein the distribution network scheduling intelligent auxiliary model takes a temperature sequence and an illumination intensity sequence as input and takes a predicted photovoltaic power generation power sequence as output.
4. The method according to claim 3, wherein in the first step, based on the historical temperature and the historical illumination intensity at each time with respect to the historical photovoltaic power at the corresponding time in the historical photovoltaic power sub-sequence, a set of dynamic influence factors is calculated by: S41, acquiring first historical temperature sub-sequences T1, T2, & gt, tm and corresponding historical illumination intensity sub-sequences L1, L2, & gt, lm, historical photovoltaic power generation power sub-sequences GP1, GP2, & gt, GPm; S42, adopting: ΔT_loc=(max(T1,T2,...,Tm)-min(T1,T2,...,Tm))/m; Calculating a local temperature change gradient delta T_loc, wherein max () is a maximum function, and min () is a minimum function; The method adopts the following steps: ΔL_loc=(max(L1,L2,...,Lm)-min(L1,L2,...,Lm))/m; Calculating a local illumination intensity change gradient delta L_loc; S43, adopting: ΔGP_T_loc=(max(GP1,GP2,...,GPm)-min(GP1,GP2,...,GPm))/(max(T1,T2,...,Tm)-min(T1,T2,...,Tm)); Calculating a temperature response gradient delta GP_T_loc; The method adopts the following steps: ΔGP_L_loc=(max(GP1,GP2,...,GPm)-min(GP1,GP2,...,GPm))/(max(L1,L2,...,Lm)-min(L1,L2,...,Lm)); calculating an illumination intensity response gradient delta GP_L_loc; S44, acquiring a historical temperature Tu corresponding to any moment in the historical temperature subsequences T1, T2, & gt, tm, wherein u is a counting index, and the value range is 1 to m; synchronously extracting historical illumination intensity Lu; s45, adopting: α_u=(ΔGP_T_loc/ΔT_loc)×(Tu/T_avg)×exp(-(Tu-T_opt) 2 /(2σ_T 2 )); calculating a temperature dynamic influence factor alpha_u, wherein T_avg is a historical temperature subsequence T1, T2, & gt, the mean value of Tm, T_opt is the optimal working temperature of the target photovoltaic array, the optimal working temperature is regarded as a known value, and sigma_T is a temperature adaptation coefficient preset by an operator in combination with the performance of the target photovoltaic array; s46, adopting: β_u=(ΔGP_L_loc/ΔL_loc)×(Lu/L_avg)×(1-exp(-Lu/L_sat)); calculating an illumination intensity dynamic influence factor beta_u, wherein L_avg is an average value of historical illumination intensity subsequences L1, L2, lm, and L_sat is an illumination saturation coefficient preset by an operator in combination with the performance of the target photovoltaic array; combining the temperature dynamic influence factor alpha_u and the illumination intensity dynamic influence factor beta_u to form a dynamic influence factor group; S47, repeating the steps, determining m temperature dynamic influence factors and m illumination intensity dynamic influence factors corresponding to m moments, and combining the m temperature dynamic influence factors and the m illumination intensity dynamic influence factors into m groups of dynamic influence factor groups according to time sequence; And summarizing m groups of dynamic influence factor groups to be used as a dynamic influence factor group set.
5. The method according to claim 4, wherein in the first step, n dynamic influence factor sets are determined, and the specific manner of constructing the training sample set as the high-dimensional environment state vector in time sequence is as follows: Repeating steps S41 to S47, and determining n dynamic influence factor group sets; extracting a kth set of dynamic influence factor groups, wherein k=1, 2, n; synchronously extracting a kth historical temperature subsequence Tk1, tk2, tkm and a kth historical illumination intensity subsequence Lk1, lk2, lkm corresponding to a kth dynamic influence factor group set; for any time u, extracting Tku and Lku-associated temperature dynamic influence factors alpha_ku and illumination intensity dynamic influence factors beta_ku; Constructing local fusion feature vectors F_ku= { T_ku, alpha_ku, L_ku and beta_ku } corresponding to the time u; Similarly, m local fusion feature vectors in the kth dynamic influence factor group set are determined, spliced in time sequence, and constructed into a high-dimensional environment state vector V_k= [ F_k1, F_k2, & gt, F_km ] associated with the kth dynamic influence factor group set as a kth input sample; Similarly, traversing n dynamic influence factor group sets, and constructing n high-dimensional environment state vectors V_1 and V_2; extracting photovoltaic power generation power values at all moments in a kth historical photovoltaic power generation power subsequence, and taking the photovoltaic power generation power values as a prediction target label Y_k corresponding to a kth input sample V_k; Combining the predicted target label Y_k with the high-dimensional environmental state vector V_k to obtain training samples (V_k, Y_k); The above steps are repeated to construct a training sample set d_tra= { (v_1, y_1), (v_2, y_2),., (v_n, y_n) }.
6. The method according to claim 5, wherein in the second step, the specific way of generating the predicted photovoltaic power sequence by predicting the photovoltaic power associated with the target photovoltaic array in the preset period in real time based on the intelligent auxiliary model for distribution network scheduling is as follows: Acquiring a preset period preset by an operator, wherein the duration of the preset period is o times; The current moment is taken as the moment before the starting moment of the preset period, a meteorological system is utilized to obtain temperature sequences T_p1, T_p2, T_po and illumination intensity sequences L_p1, L_p2, L_po corresponding to o moments of the area where the target photovoltaic array is located in the preset period; And inputting the temperature sequences T_p1, T_p2, T_po and the illumination intensity sequences L_p1, L_p2, wherein the L_po is input into a distribution network scheduling intelligent auxiliary model, and outputting corresponding predicted photovoltaic power generation power sequences GP_p1, GP_p2 and GP_po by the distribution network scheduling intelligent auxiliary model, wherein the temperature sequences and the illumination intensity sequences of the input distribution network scheduling intelligent auxiliary model are updated in real time.
7. The method according to claim 6, wherein in the third step, the specific manner of generating the power distribution network scheduling decision instruction set is: S71, obtaining a predicted photovoltaic power generation power sequence gp_p1, gp_p2. S72, acquiring predicted load sequences HP1, HP2 and HPo of the power distribution network at o times in the same preset period based on a pre-constructed load prediction system, wherein the target photovoltaic array and the power distribution network are in a grid-connected mode; s73, for any time i within a preset period, calculating a net power difference Δp_i=gp_pi-HPi of the predicted photovoltaic power generation power and the predicted load, where i=1, 2. S74, if the delta P_i-delta XP is more than or equal to 0, outputting a stable signal, enabling the target photovoltaic array to independently supply power to the power distribution network, and taking the delta P_i-delta XP as charging power to perform charging operation on the energy storage system, wherein the delta XP is redundant power generation preset by an operator; S75, if delta P_i-delta XP is less than 0, outputting a supplementary signal to enable the energy storage system to execute discharging operation, wherein the discharging power is |delta P_i-delta XP|; s76, summarizing the input signals in the steps S74 to S75 and the operation thereof, and taking the input signals as a power distribution network scheduling decision instruction corresponding to the moment i; And S77, repeating the steps, determining o power distribution network scheduling decision instructions corresponding to o moments, combining the o power distribution network scheduling decision instructions into a power distribution network scheduling decision instruction set according to time sequence, transmitting the power distribution network scheduling decision instruction set to an operator for secondary regulation and control, and determining a regulation and control instruction by the operator.
8. The method according to claim 7, wherein in the fourth step, the specific manner of determining the issuing operation of the power distribution network scheduling decision instruction set is: extracting a regulation instruction of an operator, and checking; If the regulation instruction is a issuing instruction, extracting a power distribution network scheduling decision instruction set without changing, and executing issuing operation to the power distribution network; And if the regulation and control instruction is a change instruction, extracting a power distribution network scheduling decision instruction set after the change of an operator, and executing issuing operation to the power distribution network.
9. The method of claim 8, wherein in the fourth step, the specific manner of performing feedback optimization on the intelligent auxiliary model for distribution network scheduling is: extracting the actual illumination intensity and the actual temperature at o times in a preset period to form an actual illumination intensity sequence and an actual temperature sequence; Synchronously extracting actual photovoltaic power generation power of the photovoltaic power generation array at o times to form an actual photovoltaic power generation power sequence; And generating a feedback sample set based on the contents in the steps S34 to S37 by using the actual illumination intensity sequence, the actual temperature sequence and the actual photovoltaic power generation power sequence, and performing feedback optimization on the intelligent auxiliary model of the distribution network scheduling.

Description

Intelligent scheduling decision method based on intelligent auxiliary distribution network scheduling model Technical Field The invention belongs to the technical field of distribution network scheduling, and particularly relates to an intelligent scheduling decision method based on an intelligent auxiliary distribution network scheduling model. Background Photovoltaic is taken as a clean energy source, plays an increasingly important role in energy supply, and an intelligent auxiliary model is applied to distribution network scheduling so as to realize the optimization of precise prediction and scheduling decision of photovoltaic power generation power, and has important significance for improving the operation efficiency of a power distribution network and the digestion capability of renewable energy sources. In the prior art, when a power distribution network with high-proportion photovoltaic access is processed, the power distribution network is mostly dependent on a relatively simplified prediction model and static experience rules, and obvious defects and drawbacks exist; A statistical prediction method based on historical power generation data or a simple physical formula model is often adopted, the model input is usually only a rough historical power curve or a single and lagged meteorological observation value, the complex dynamic relevance and space-time coupling effect between key environment parameters such as illumination intensity, temperature and the like and output power of a photovoltaic array cannot be deeply excavated and quantized, prediction precision is reduced when the model is used for coping with rapid fluctuation scenes such as weather shock and cloud layer movement, short-time and high-frequency fluctuation characteristics of photovoltaic power generation cannot be accurately captured, in a scheduling decision, a prediction-planned offline mode or a simple start-stop rule based on a fixed threshold value is mostly adopted in the prior art, on-line optimization matching capability linked with a high-precision ultra-short-term prediction result is lacked, scheduling instructions are enabled to generate stiffness, instantaneous differences between photovoltaic output and load demands cannot be finely stabilized, or renewable energy waste is caused, or a traditional standby unit is required to be frequently adjusted, and running cost and equipment loss are increased. In order to solve the problems, the invention provides an intelligent scheduling decision method based on an intelligent auxiliary model of distribution network scheduling. Disclosure of Invention Aiming at the defects of the prior art, the invention provides an intelligent scheduling decision method based on an intelligent auxiliary scheduling model of a distribution network, which solves the problems of the prior art that the photovoltaic power prediction precision is insufficient, the scheduling decision instantaneity is poor and the self-adaption capability is lacking. The aim of the invention can be achieved by the following technical scheme: An intelligent scheduling decision method based on a distribution network scheduling intelligent auxiliary model, the method comprises the following steps: Step one, obtaining photovoltaic power generation power related to a target photovoltaic array and a time sequence from a cloud database; synchronously extracting illumination intensity and temperature, and constructing a distribution network scheduling intelligent auxiliary model related to the environment of the target photovoltaic array; Acquiring the illumination intensity and the temperature associated with the area where the target photovoltaic array is located in a preset period in real time based on a meteorological system, and predicting the photovoltaic power generation power associated with the target photovoltaic array in the preset period in real time based on an intelligent auxiliary model for distribution network scheduling to generate a predicted photovoltaic power generation power sequence; optimizing and matching based on the predicted photovoltaic power generation power sequence and a predicted load sequence associated with the power distribution network, and generating a power distribution network scheduling decision instruction set; And step four, determining issuing operation of a power distribution network scheduling decision instruction set based on an operator regulation instruction, extracting actual illumination intensity and actual temperature associated with time sequence in a preset period, generating a feedback sample set together with actual photovoltaic power generation power, and performing feedback optimization on an intelligent auxiliary model of power distribution network scheduling. As a further scheme of the present invention, in the first step, a specific manner of obtaining the photovoltaic power generation power associated with the target photovoltaic array and the time sequence from