Search

CN-121981701-A - Self-adaptive cleaning decision method and system for photovoltaic power station under resource unconstrained scene

CN121981701ACN 121981701 ACN121981701 ACN 121981701ACN-121981701-A

Abstract

The invention discloses a self-adaptive cleaning decision method and system for a photovoltaic power station under a resource unconstrained scene. The invention aims at a resource unconstrained scene of a machine, and comprises the steps of carrying out distributed sensing by using a single photovoltaic array and a dedicated cleaning robot thereof as independent units, constructing an independent profit-cost dynamic starting threshold model for each unit, calculating a net cleaning utility value based on a refined weather forecast when the pollution rate exceeds a threshold, immediately triggering cleaning if the pollution rate is smaller than a mud spot risk threshold, executing a rainwater replacement strategy if the pollution rate is larger than the cleaning cost, combining future time-sharing electricity price curve optimizing when the cleaning is needed, selecting a moment with the maximum expected net profit, starting operation, and carrying out closed-loop feedback and iterative updating on a model reference coefficient based on actual energy efficiency lifting data after the cleaning is completed. According to the invention, intelligent and self-adaptive optimization of cleaning decision is realized by introducing rainfall effectiveness dynamic game, electricity price window active optimizing and online self-learning correction mechanisms.

Inventors

  • SUN SHAOYUN
  • CHU YI
  • XU JIANJIANG
  • YUAN YE

Assignees

  • 江苏国信泗阳太阳能发电有限公司
  • 江苏省新能源开发股份有限公司

Dates

Publication Date
20260505
Application Date
20251225

Claims (10)

  1. 1. The self-adaptive cleaning decision method of the photovoltaic power station under the resource unconstrained scene is characterized by comprising the following steps of: s1, a single photovoltaic array and a dedicated cleaning robot are taken as independent decision units, and real-time pollution rate data, weather forecast data, time-of-use electricity price information and equipment state data of the cleaning robot of the unit are collected in real time; s2, calculating a dynamic pollution rate threshold value for the independent decision unit, comparing the real-time pollution rate with the dynamic pollution rate threshold value, and deciding to keep standby if the real-time pollution rate does not exceed the dynamic pollution rate threshold value; s3, if the real-time pollution rate exceeds the dynamic pollution rate threshold, weather forecast in a future preset time window is obtained, a net cleaning utility value of forecast rainfall is calculated, and when the net cleaning utility value is smaller than a preset mud spot risk threshold, an instruction for immediately executing machine cleaning is generated; S4, after determining that machine cleaning is required to be executed, calculating expected net benefits of immediate cleaning and expected net benefits of cleaning delayed to all potential low-electricity-price time periods in the future by combining a time-of-use electricity price curve of the future for N hours, selecting the moment with the maximum expected net benefits as a final operation starting point, and generating corresponding cleaning instructions; And S5, after the cleaning operation is finished, calculating the actual energy efficiency improvement gain before and after the cleaning of the array based on the actual operation data fed back by the exclusive cleaning robot, calculating the deviation between the actual energy efficiency improvement gain and the cleaning gain estimated when the cleaning decision is made, and carrying out iterative updating on the reference coefficient in the dynamic pollution rate threshold value through an online learning algorithm according to the deviation, wherein the updated reference coefficient is used for the next decision cycle of the independent decision unit.
  2. 2. The method for adaptively cleaning and deciding a photovoltaic power station in a resource unconstrained scenario according to claim 1, wherein the calculation formula of the dynamic pollution rate threshold θ S(t) is as follows: Wherein k 0 is a reference coefficient updated through online learning, OC i is a single cleaning operation cost of an ith photovoltaic array, P_avg (t) is a predicted average electricity price in a future cleaning effect period, P_ irated is rated power of the photovoltaic array, k_ season is a season adjustment coefficient set based on illumination resource abundance of a position where the photovoltaic array is located, and k_weather is a weather adjustment coefficient set based on weather forecast continuity of the photovoltaic array.
  3. 3. The method for adaptive cleaning decision of photovoltaic power plants in a resource-unconstrained scenario according to claim 2, wherein the seasonal adjustment factor k season takes a negative value in summer to encourage advanced cleaning and takes a positive value in winter to raise the threshold.
  4. 4. The method for adaptively cleaning and deciding a photovoltaic power plant in a resource unconstrained scenario according to claim 2, wherein the weather adjustment coefficient k_weather takes a negative value to decrease the threshold value when forecasting continuous sunny days and takes a positive value to increase the threshold value when forecasting intermittent dust or rainy days.
  5. 5. The method for adaptive cleaning decision-making of a photovoltaic power plant in a resource-unconstrained scenario according to any one of claims 2 to 4, wherein in step S3: When the predicted rainfall is lower than a first threshold, determining that the net cleaning utility value is negative; When the forecast rainfall amount is higher than the second threshold value, determining that the net cleaning utility value is positive; when the predicted rainfall is between the first threshold and the second threshold, the plaque effect probability is evaluated based on the historical data model and a net cleaning utility value is calculated.
  6. 6. The method for adaptively cleaning and deciding a photovoltaic power plant in an unconstrained resource scenario according to claim 5, wherein in step S4, the future N hours are 24 hours, and the expected net benefit calculation formula is expected net benefit= (expected gain power generation amount after cleaning x corresponding period power rate) -cleaning cost-loss of power generation due to pollution during delay.
  7. 7. The self-adaptive cleaning decision method for the photovoltaic power station in the resource unconstrained scene according to claim 1, wherein the weather forecast data are lattice weather data containing rainfall forecast within 72 hours, and the time-of-use electricity price information is obtained in real time through an electric network power marketing system interface.
  8. 8. The utility model provides a photovoltaic power plant's self-adaptation washs decision-making system under resource unconstrained scene which characterized in that includes: The sensing module is used for collecting pollution rate data, weather forecast data, time-of-use electricity price information and equipment state data of the corresponding exclusive cleaning robot of each independent decision unit in real time by taking a single photovoltaic array as a unit; the decision module is in communication connection with the perception module and is deployed at an edge computing node close to the photovoltaic array, and is used for independently executing steps S2-S5 in the method of any one of claims 1-7 for each independent decision unit to generate a cleaning instruction; And the execution module is in communication connection with the decision module and is used for receiving the cleaning instruction, scheduling the corresponding exclusive cleaning robot to execute cleaning operation and feeding back an operation result to the sensing module to form a closed loop.
  9. 9. An electronic device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, characterized in that the processor implements the steps of the method according to any of claims 1-7 when the program is executed.
  10. 10. A non-transitory computer readable storage medium, on which a computer program is stored, characterized in that the computer program, when being executed by a processor, implements the steps of the method according to any of claims 1-7.

Description

Self-adaptive cleaning decision method and system for photovoltaic power station under resource unconstrained scene Technical Field The invention relates to the technical field of operation and maintenance of photovoltaic power stations, in particular to a photovoltaic power station self-adaptive cleaning decision method and system based on dynamic threshold, multi-factor game and closed-loop learning under a resource unconstrained (namely 'one machine gust') scene. Background The installed capacity of photovoltaic power generation is continuously increased under the promotion of 'carbon peak, carbon neutralization' strategic targets. However, the deposition of pollutants such as dust, bird droppings and the like on the surface of the photovoltaic module can cause serious shielding effect, and researches show that the pollution can reduce the power generation efficiency of the power station by 17-40%, so that the economic benefit is seriously influenced. Currently, intelligent cleaning decision research of photovoltaic power plants is mostly based on "resource-constrained" scenarios, i.e. limited cleaning equipment, such as mobile robots or cleaning teams, need to be scheduled among a large number of photovoltaic arrays. Such problems are typically modeled as a traveler problem or a vehicle path problem and solved using heuristic algorithms such as particle swarm optimization, genetic algorithms, and the like. Although the methods are feasible in theory, the problems of high calculation complexity, sensitive model parameters, weak real-time decision making capability and the like are faced in the actual engineering, and the large-scale application of the methods is limited. In recent years, with the popularity of the "one-machine-array" mode, resource constraints have been released, and the decision core has turned to finding a cleaning opportunity for each individual array that is globally profitable. However, the existing schemes are mostly simple rules, such as a fixed period, a single pollution rate threshold value or a static multi-factor combination, and have obvious defects that (1) the 'double-blade sword' effect of natural rainfall cannot be accurately quantified, the 'mud spot effect' is possibly caused by small rain, the generated energy is reduced instead, (2) dynamic benefit optimization cannot be carried out with time-sharing electricity price, only high price period is simply avoided, and higher benefit opportunities are possibly missed, (3) self-learning and correction capability is lacked, model parameters are fixed, and array characteristic differences and environment long-term changes cannot be adapted. Thus, there is a need for an adaptive decision scheme that enables dynamic gaming, revenue optimization, and self-evolution. Disclosure of Invention The invention aims to overcome the defects of the prior art and provides a self-adaptive cleaning decision method and a self-adaptive cleaning decision system for a photovoltaic power station under a resource unconstrained scene. The invention not only realizes the distributed independent decision, but also solves the problems of inaccurate cleaning decision, poor economy and lack of self-adaptability in the prior art by introducing an effective rainfall game mechanism, an electricity price window profit optimizing model and a closed-loop correction mechanism based on actual effects. In a first aspect, the invention provides a self-adaptive cleaning decision method for a photovoltaic power station in a resource unconstrained scene, which comprises the following steps: S1, distributed sensing and data acquisition, namely, using a single photovoltaic array and a dedicated cleaning robot thereof as an independent decision unit, and acquiring real-time pollution rate data, weather forecast data, time-of-use electricity price information and equipment state data of the cleaning robot in real time; S2, calculating a dynamic pollution rate threshold value theta S(t) for the independent decision unit, comparing the real-time pollution rate with the dynamic pollution rate threshold value theta S(t), and deciding to keep standby if the real-time pollution rate does not exceed the dynamic pollution rate threshold value; And S3, effective rainfall game judgment, namely if the real-time pollution rate exceeds a dynamic pollution rate threshold value theta S(t), acquiring a refined weather forecast (comprising rainfall probability and forecast rainfall) within a preset time window (such as 24-48 hours) in the future, and calculating a net cleaning utility value of the forecast rainfall, wherein the calculation is characterized in that the quantitative evaluation of the risk of the soil spot effect caused by the rainfall is performed. When the net cleaning utility value is smaller than a preset mud spot risk threshold value, determining that rainfall waiting is uneconomic or risk exists, ignoring rainfall factors, and generating an instruction for immediatel