Underground garage illumination control system and method based on artificial neural network PID, and storage medium

Abstract

The invention relates to the technical field of intelligent lighting control systems, in particular to an underground garage lighting control system and method based on artificial neural network PID and a storage medium. Comprising a brightness sensor, a traffic flow sensor, an illuminance sensor, a temperature and humidity sensor and a microwave speed measurement sensor which are arranged inside and outside a garage; the intelligent network connection lamp comprises an intelligent gateway unit, an LED dimming driving unit and an LED lamp, and the intelligent gateway unit is used for receiving the control instruction and communicating with the upper computer; and the upper computer is used for receiving the data of the detection unit, generating an optimal illumination control curve based on a garage comfort illumination model and a rendezvous zone fuzzy control model, and adjusting the illumination of lamps in each zone in real time through an RBF neural network PID (Proportion Integration Differentiation) algorithm. According to the method, self-learning optimization is achieved, PID parameters are adjusted online through the neural network RBF, the method adapts to long-term dynamic influences such as differences of different types of garages, seasonal changes and equipment aging, and optimal control is given.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent lighting control systems, and particularly to an underground garage lighting control system, method and storage medium based on an artificial neural network PID. Background Art

[0002] Underground garages are places where people often enter and exit in daily life. Statistical analysis shows that the electricity consumption for garage lighting accounts for 40% of the total electricity consumption of the underground garage. How to control garage lighting comfortably, efficiently and greenly has become an urgent problem to be solved.

[0003] The existing underground garage lighting controls mainly include constant lighting, timing, radar induction, network control, etc. The illuminance in the garage is affected by many factors. The illuminance has a direct relationship with the wall reflectance, ground reflectance, ceiling reflectance, etc. The cleanliness of the garage, the wiping frequency of the lamps, and the light decay due to lamp aging will all directly affect the output illuminance. The existing controls have set the control parameters at the factory and cannot dynamically perceive the changes in the objective environment of the garage and factors such as lamp aging. Therefore, the output illuminance cannot meet people's real illuminance requirements in real time.

[0004] The existing technologies have the following problems in the actual application process:

[0005] For the constant lighting control of garage lighting, no control is performed on the garage lighting, resulting in energy waste. Since the lamps are turned on for 24 hours, the lamp life is reduced.

[0006] For the timing control of garage lighting, the lighting in the garage area is turned on or off according to a certain time period. Although it can save energy to some extent, the overall effect is not comfortable. Ergonomics (human eye adaptation) is not considered, which is likely to cause vehicle accidents. After the lighting is turned off, it poses a serious threat to the safety of pedestrians and vehicles.

[0007] For the radar induction control, the lights turn on when a vehicle comes and turn off when the vehicle leaves. The radar induction has a range limitation. The lights within 10 meters in front of the vehicle are turned on, achieving partial energy saving. The group control of the lights is not realized, resulting in poor illuminance in the front of the field of vision (more than 10 meters away). The place with a vehicle is bright and the place without a vehicle is dark. The illuminance uniformity of the entire garage is poor, which is likely to cause vehicle collision accidents.

[0008] For the network control, the group control of the lamps is realized, which is basically timing control + illuminance adjustment. The illuminance adjustment reduces a certain value by default. The adjustment process does not consider the influence of seasonal changes and vehicle flow changes on the garage illuminance, nor does it consider the comfort of the human eye. It is impossible to achieve lighting on demand and comfortable lighting.

[0009] Therefore, the present invention proposes an underground garage lighting control system, method and storage medium based on an artificial neural network PID. Summary of the Invention

[0010] To solve at least one of the above-mentioned technical problems, the present invention proposes an underground garage lighting control system, method and storage medium based on an artificial neural network PID.

[0011] The present invention is realized through the following technical solutions: An underground garage lighting control system based on an artificial neural network PID divides the underground garage into an entrance section, a transition section, a basic section and an intersection area, and is characterized by including:

[0012] A detection unit, including brightness sensors, traffic flow sensors, illuminance sensors, temperature and humidity sensors, and microwave speed sensors arranged inside and outside the garage;

[0013] An intelligent connected light, including an intelligent gateway unit, an LED dimming drive unit and an LED lamp, and the intelligent gateway unit is used to receive control instructions and communicate with the upper computer;

[0014] An upper computer, which is used to receive the data of the detection unit, generate an optimal illuminance control curve based on the garage comfort illuminance model and the intersection area fuzzy control model, and adjust the illuminance of the lamps in each area in real time through the RBF neural network PID algorithm.

[0015] Further, one brightness sensor is arranged inside and outside the garage entrance and exit;

[0016] The illuminance sensors are arranged at the garage entrance, the garage ramp, the middle of the garage, and the garage intersection area;

[0017] One temperature and humidity sensor is arranged inside and outside the garage;

[0018] One traffic flow sensor is arranged in each of the four directions of the intersection area;

[0019] One microwave speed sensor is arranged in each of the four directions of the intersection area

[0020] An underground garage lighting control method includes the following method steps:

[0021] Collect the traffic flow in the intersection area, collect the brightness inside and outside the garage, and collect the illuminance in the entrance section, transition section, basic section and intersection area of the garage;

[0022] Calculate the illuminance reference values of the entrance section, transition section and basic section according to the garage comfort illuminance model;

[0023] Combined with the current traffic flow, give the illuminance reference value of the intersection area according to fuzzy control;

[0024] Draw the optimal illuminance curve of the garage;

[0025] Using the given optimal illuminance curve, the RBF neural network PID is used to control the lamp drive until the optimal illuminance is achieved.

[0026] Further, the garage comfort illuminance model includes algorithm models corresponding to the entrance section, transition section, and basic section, where:

[0027] Algorithm model of the entrance section: The entrance section is set in the entrance area of the garage. The illuminance reference for the entrance section is defined as follows:

[0028]

[0029] L out Outside the garage; Luminance; k reduction coefficient, with a value of 0.8;

[0030] Algorithm model of the transition section: The transition section is set in the ramp area where the vehicle enters the garage. The illuminance reference for the transition section is defined as follows:

[0031]

[0032] L in Luminance inside the garage; L out Luminance outside the garage; X Distance of the vehicle from the starting point of the transition section; L total Length of the transition section;

[0033] Algorithm model of the basic section: The basic section is located inside the garage. Taking the vehicle flow Q as a variable, a dynamic illuminance adjustment formula is proposed:

[0034]

[0035] Eb: Basic illuminance, taking 50 LX for public garages and 30 LX for residential garages according to national standards; Q: Measured vehicle flow, number of vehicles passing through per unit time; Q max : Maximum vehicle flow; α: Adjustment coefficient;

[0036] Calibration method of the adjustment coefficient α: Optimize the α value in real time by combining the Kalman filter:

[0037] α t =α t-1 +K(Q n -Q n-1 );

[0038] Among them, K is the Kalman gain coefficient;

[0039] Nonlinear relationship modeling: When the vehicle flow exceeds the threshold, the illuminance demand shows an exponential growth. A piecewise function is introduced:

[0040]

[0041] Furthermore, the rule base of the fuzzy control includes the following rules:

[0042] If the traffic flow is large and the illumination in the basic section is high, then the illumination in the intersection area is set to high;

[0043] If the traffic flow is large and the illumination in the basic section is low, then the illumination in the intersection area is set to high;

[0044] If the traffic flow is small and the illumination in the basic section is low, then the illumination in the intersection area is set to low.

[0045] Furthermore, the inference of the fuzzy control includes the following steps:

[0046] Fuzzification: The traffic flow detection value Q(K) and the illumination E(K) in the basic section are used as the inputs of the fuzzy controller, and the output of the fuzzy controller is the illumination value u in the intersection area. Define the subsets of the traffic flow Q(K), the illumination E(K) in the basic section, and the reference illumination value u in the intersection area as {small, medium, large}, and the universe of discourse is [0, 1]; The triangular membership function is adopted.

[0047] Fuzzy inference: The fuzzy inference adopts Mamdani inference, and the inference algorithm is as follows:

[0048] η i = μ ei [e(k)] ∧ μ eci [ec(k)] = min(μ ei [e(k)] · μ eci [ec(k)]);

[0049] η i is the fuzzy inference calculation value; μ ei [e(k)] is the traffic flow membership value; μ eci [ec(k)] is the illumination membership value in the basic section; ∧ is the minimum operation; The fuzzy rule base is established.

[0050] Defuzzification: The defuzzification adopts the centroid method; The defuzzification calculation method is as follows:

[0051]

[0052] Furthermore, the RBF neural network PID control of the lamp drive includes the following steps:

[0053] Receive the optimal illumination curve;

[0054] Collect the current illumination and calculate the error;

[0055] Real-time adjust the lamp illuminations in each area through the RBF artificial neural network PID algorithm.

[0056] Furthermore, the RBF artificial neural network PID algorithm includes:

[0057] The structure of the RBF neural network. Suppose the network is a three-layer feedforward structure:

[0058] Input layer (2 nodes):

[0059]

[0060] Hidden layer (m nodes):

[0061] Gaussian radial basis function:

[0062] Where:

[0063] The center vector of the j-th hidden node; σ j : The width of the j-th hidden node;

[0064] Output layer (3 nodes):

[0065] PID parameters:

[0066] Where w pj , w ij , w dj are the output weights, and b p , b i , b d are the bias terms;

[0067] PID control law, dynamic dimming output:

[0068]

[0069] u(t) is the output of the controller;

[0070] e(t) is the error signal (the difference between the set value and the actual value);

[0071] Kp, Ki, and Kd are the proportional, integral, and derivative gains respectively.

[0072] The present invention also provides a computer-readable storage medium storing a computer program, and when the program is executed by a processor, it implements the underground garage lighting control method described above.

[0073] Compared with the prior art, the present invention has the following beneficial effects:

[0074] 1. People-oriented and lighting on demand. After studying the mechanism of human eye dark adaptation, a comfort illuminance model for the garage is established, and the optimal illuminance reference for the garage is given.

[0075] 2. Safer. The working mechanism of the garage intersection area is studied, and a fuzzy controller is established with traffic flow and basic garage illumination as inputs and intersection area illumination reference as output. With the help of expert experience, complex control is simplified and made more efficient. The dynamically adjusted intersection area illumination improves the safety of garage operation on the one hand and reduces power consumption on the other hand.

[0076] 3. According to the optimal illumination curve of the garage generated by the garage comfort illumination model and the fuzzy control of the intersection area, the RBF artificial neural network PID algorithm is used to adjust the illumination of lamps in each area in real time to achieve closed-loop tracking control.

[0077] 4. It achieves lighting on demand. When the traffic flow is large and the brightness difference between inside and outside the garage is large and the illumination needs to be increased, high illumination is output. When the traffic flow is small, there are few people passing by, and at night, low illumination is output. While ensuring comfort and safety, it greatly reduces the power consumption of the garage.

[0078] 5. The RBF-PID control of the artificial neural network in this algorithm has the following control advantages, thus achieving more intelligent lighting control

[0079] (1) Nonlinear adaptation ability: Approximating complex mappings through Gaussian kernel functions to solve the dynamic coupling problem between traffic flow and illumination.

[0080] (2) Self-learning optimization: Online adjusting PID parameters to adapt to long-term dynamic influences such as differences in different types of garages, seasonal changes, and equipment aging.

[0081] (3) Anti-interference ability: In low-light or sensor noise environments, its robustness is better than that of traditional PID;

[0082] 6. Eliminate objective interference factors. The garage illumination is directly related to the wall reflectance, floor reflectance, cleanliness of the garage, cleanliness of the lamps, lamp aging, etc. Traditional dimming control cannot consider these objective factors. This algorithm self-learns and optimizes, uses the neural network RBF to online adjust PID parameters, adapts to long-term dynamic influences such as differences in different types of garages, seasonal changes, and equipment aging, and gives the optimal control. Description of the Drawings

[0083] Figure 1 It is the overall system composition diagram of the present invention.

[0084] Figure 2 It is the overall system working flow chart of the present invention.

[0085] Figure 3 It is the functional partition diagram of the underground garage of the present invention.

[0086] Figure 4 It is the optimal illumination curve diagram of the underground garage of the present invention.

[0087] Figure 5 It is the membership function curve graph of the traffic flow Q(K) of the present invention.

[0088] Figure 6 It is the membership function curve graph of the basic section illuminance E(K) of the present invention.

[0089] Figure 7 It is the membership function curve graph of the illuminance reference value u of the present invention.

[0090] Figure 8 It is the fuzzy control flow chart of the present invention.

[0091] Figure 9 It is the logic block diagram of the RBF network PID control of the present invention.

[0092] Figure 10 It is the RBF network PID control flow chart of the present invention.

[0093] Figure 11 It is the schematic diagram of the rules of the fuzzy controller built by the present invention.

[0094] Figure 12 It is the schematic diagram of the defuzzification process of the present invention.

[0095] Figure 13 It is the simulation curve graph of the RBF neural network PID control of the present invention. Specific implementation manners

[0096] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0097] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by those of ordinary skill in the technical field to which the present invention belongs. The terms used in the description of the present invention herein are only for the purpose of describing specific embodiments, and are not intended to limit the present invention. The term "or / and" used herein includes any and all combinations of one or more of the related listed items.

[0098] Please refer to Figures 1-10 , this embodiment provides an underground garage lighting control system based on an artificial neural network PID. The system is composed of a detection unit, intelligent networked lights, and a host computer.

[0099] (1) The detection unit includes a brightness sensor, a traffic flow sensor, an illuminance sensor, a temperature and humidity sensor, a traffic flow sensor, and a microwave speed measurement sensor.

[0100] (1) One brightness sensor is arranged inside and one outside the garage entrance and exit.

[0101] (2) The illuminance sensors are arranged at the garage entrance, the garage ramp, the middle of the garage, and the garage intersection area.

[0102] (3) One temperature and humidity sensor is arranged inside the garage and one outside the garage.

[0103] (4) One traffic flow sensor is arranged in each of the four directions in the intersection area.

[0104] (5) One microwave speed measurement sensor is arranged in each of the four directions in the intersection area.

[0105] (2) The intelligent connected light consists of an intelligent gateway unit, an LED dimming drive unit, and an LED light.

[0106] The intelligent gateway unit is responsible for control and communication. On the one hand, it is responsible for receiving instructions from the upper computer, and on the other hand, it is responsible for wireless or wired networking communication with other lamps and the upper computer.

[0107] The LED dimming drive unit is responsible for dimming according to the instructions of the upper computer.

[0108] (3) The upper computer mainly realizes the control of the entire system.

[0109] As Figure 1 shown, the upper computer is the control center of the entire system. It receives the data collection of the detection unit, analyzes and processes the data, calculates through the garage comfort illuminance model and the intersection area illuminance model, gives the optimal illuminance control curve of the garage, and sends it to the intelligent connected lights in each area.

[0110] As Figure 2 shown, the present invention also proposes an underground garage lighting control method, including the following method steps:

[0111] (1) The system detects the brightness difference inside and outside the garage, detects the traffic flow, and detects the illuminance inside the garage (entrance section, transition section, basic section, intersection area).

[0112] (2) Calculate the illuminance references for the entrance section, transition section, and basic section according to the garage comfort illuminance model.

[0113] (3) Combine the current traffic flow and give the illuminance reference for the intersection area according to fuzzy control.

[0114] (4) Draw the optimal illuminance control curve of the garage.

[0115] 5) Utilize the given optimal illuminance curve and use RBF neural network PID to control the lamp drive until the optimal illuminance is achieved.

[0116] Construction of the comfort illuminance model for the garage:

[0117] When the human eye enters the dark environment inside the garage from the bright environment outside the garage, an adaptation process is required, which is dark adaptation. Dark adaptation is divided into two stages:

[0118] (1) Rapid adaptation stage. The retinal cone cells are responsible for brightness adjustment and it takes 5 minutes.

[0119] (2) Complete adaptation stage. The rod cells are gradually activated and it takes 30 minutes.

[0120] As Figure 3 shown, Figure 3 Figure 3 is the functional zoning diagram of the underground garage. To ensure visual comfort inside the garage, the garage needs to be functionally divided into an entrance section, a transition section, and a basic section. Using the adaptive dimming algorithm, the illuminance inside the garage is gradually changed in brightness to meet the adaptation requirements of the human eye and provide comfortable lighting.

[0121] (1) Entrance section, which is set in the entrance area of the garage. The illuminance reference for the entrance section is defined as follows:

[0122]

[0123] L out Brightness outside the garage;

[0124] k reduction coefficient, with a value of 0.8;

[0125] (2) Transition section, which is set in the ramp area where the vehicle enters the garage. The illuminance reference for the transition section is defined as follows:

[0126] Achieve smooth brightness transition through the cosine interpolation function, and dynamically adjust the length of the transition section in combination with the vehicle speed, which can meet the adaptation requirements of the human eye. During actual deployment, the parameters need to be calibrated according to the measured data, and the PID or fuzzy control algorithm should be embedded to cope with environmental disturbances (such as sudden changes in light during rainy or cloudy weather).

[0127]

[0128] L in Brightness inside the garage;

[0129] L out Brightness outside the garage;

[0130] X Distance of the vehicle from the starting point of the transition section;

[0131] L total Length of the transition section;

[0132] (3) The basic section is located inside the garage. The derivation of the reference value of the illuminance in the basic section is as follows:

[0133] 1. Basic illuminance calculation model:

[0134] According to the utilization factor method in the "Standard for Architectural Lighting Design", the formula for average illuminance is:

[0135]

[0136] Where:

[0137] E: Average illuminance (lx);

[0138] N: Number of luminaires;

[0139] Φ: Luminous flux of a single luminaire (lm);

[0140] U: Utilization factor (related to luminaire layout and reflectivity);

[0141] K: Maintenance factor (usually taken as 0.7 - 0.8);

[0142] A: Area of the lighting area m 2 .

[0143] 2. Dynamic dimming model for vehicle flow:

[0144] Taking the vehicle flow Q (number of vehicles passing through per unit time) as a variable, a dynamic illuminance adjustment formula is proposed:

[0145]

[0146] E b : Basic illuminance (taking 50 LX for public garages and 30 LX for residential garages according to national standards);

[0147] Q: Measured vehicle flow;

[0148] Q max : Maximum vehicle flow;

[0149] α: Adjustment coefficient;

[0150] Calibration method for the adjustment coefficient α: Combining the Kalman Filter to optimize the α value in real time:

[0151] α t = α t-1 + K(Q n - Q n-1 );

[0152] Where K is the Kalman gain coefficient.

[0153] Non - linear relationship modeling: When the traffic flow exceeds the threshold, the illuminance demand increases exponentially. A piece - wise function is introduced:

[0154]

[0155] According to the above - mentioned garage comfort illuminance model, the optimal illuminance curve at the current moment is as Figure 4 shown.

[0156] This curve fully considers the changes in the brightness difference between inside and outside the garage (such as factors like sunny days, cloudy days, seasonal changes, weather changes, etc.) and the comfort of the human eye, achieving people - oriented, lighting on - demand, and energy - saving lighting.

[0157] Illuminance adjustment in the intersection area based on traffic flow prediction:

[0158] The present invention uses fuzzy intelligent control, taking the traffic flow in the intersection area and the illuminance of the basic section (calculated by the garage comfort illuminance model) as the inputs of the fuzzy logic controller, and the output is the illuminance of the intersection area.

[0159] The basic idea is:

[0160] If the illuminance of the basic section calculated according to the garage comfort illuminance model is relatively high and the traffic flow is large, it indicates that it is daytime and the peak period of garage use, and the illuminance of the intersection area needs to be increased.

[0161] If the illuminance of the basic section calculated according to the garage comfort illuminance model is relatively low and the traffic flow is large, it indicates that it is night (or cloudy day) and the peak period of garage use, and the illuminance of the intersection area needs to be increased.

[0162] If the illuminance of the basic section calculated according to the garage comfort illuminance model is relatively low and the traffic flow is low, it indicates that it is night (or cloudy day) and the low - peak period of garage use, and the illuminance of the intersection area needs to be decreased.

[0163] Fuzzy controller design:

[0164] Based on the above three basic ideas, a fuzzy controller for the illuminance of the garage intersection area is designed. Fuzzy control reasoning mainly includes three processes: namely, fuzzification, fuzzy reasoning, and defuzzification.

[0165] 1. Fuzzification

[0166] The present invention selects the traffic flow detection value Q(K) and the illuminance of the basic section E(K) as the inputs of the fuzzy controller, and the output of the fuzzy controller is the illuminance value u of the intersection area.

[0167] Define the subsets of the traffic flow Q(K), the illuminance of the basic section E(K), and the reference illuminance value u of the intersection area as {small, medium, large}, and the universe of discourse is [0,1].

[0168] The triangular membership function is adopted. The membership function curve of the traffic flow Q(K) is as shown in Figure 5 Figure Figure 6 shown, and the membership function curve of the basic section illuminance E(K) is as shown in Figure 7 Figure

[0169] 2. Fuzzy inference:

[0170] Mamdani inference is adopted for fuzzy inference.

[0171] (1) The adopted inference algorithm is:

[0172] η i = μ ei [e(k)] ∧ μ eci [ec(k)] = min(μ ei [e(k)] · μ eci [ec(k)])

[0173] η i is the calculated value of fuzzy inference; μ ei [e(k)] is the membership value of traffic flow; μ eci [ec(k)] is the membership value of the basic section illuminance; ∧ is the minimum operation.

[0174] Establishment of the fuzzy rule base:

[0175] When designing the fuzzy rules, the present invention deeply studied the characteristics of garage operation and set up a fuzzy rule base based on actual operation. An example of the idea for establishing the fuzzy rule base is as follows:

[0176] (1) If the basic section illuminance calculated according to the garage comfort illuminance model is relatively high and the traffic flow is large, it indicates that it is daytime and the peak period of garage use, and the illuminance of the intersection area needs to be increased.

[0177] The corresponding rule design is:

[0178] IF Q is large and E is large, then u is L.

[0179] (2) If the basic section illuminance calculated according to the garage comfort illuminance model is relatively low and the traffic flow is large, it indicates that it is night (or cloudy) and the peak period of garage use, and the illuminance of the intersection area needs to be increased.

[0180] The corresponding rule design is:

[0181] IF Q is large and E is small, then u is L.

[0182] (3) If the illuminance of the basic section calculated according to the garage comfort illuminance model is low and the traffic flow is low, it indicates that it is at night (or on a cloudy day) and the low-usage period of the garage, and the illuminance of the intersection area needs to be reduced.

[0183] The corresponding rule design is as follows:

[0184] IF Q is small and E is small, then u is S.

[0185] According to the above design idea, a total of the following 9 fuzzy rules are designed.

[0186] IF Q is small and E is small, then u is small.

[0187] IF Q is small and E is medium, then u is small.

[0188] IF Q is small and E is large, then u is small.

[0189] IF Q is medium and E is small, then u is medium.

[0190] IF Q is medium and E is medium, then u is medium.

[0191] IF Q is medium and E is large, then u is medium.

[0192] IF Q is large and E is small, then u is large.

[0193] IF Q is large and E is medium, then u is large.

[0194] IF Q is large and E is large, then u is large.

[0195] The meaning of the formula in the rule:

[0196] Q: Traffic flow;

[0197] E: Illuminance of the basic section;

[0198] u: Reference illuminance;

[0199] S: Small;

[0200] M: Medium;

[0201] L: Large;

[0202] 3. Defuzzification:

[0203] The centroid method is used for defuzzification. The defuzzification calculation method is as follows:

[0204]

[0205] The overall working flowchart of the fuzzy algorithm is as Figure 8 shown.

[0206] Figure 5 Let Q(k) be the input of the fuzzy controller;

[0207] Figure 6 Let E(k) be the input of the fuzzy controller;

[0208] Figure 7 Let u be the output of the fuzzy controller;

[0209] PID illuminance control based on RBF network:

[0210] Each lamp is equipped with an address code. According to the illuminance value reference calculated by the host computer, the intelligent controller is used to adjust the LED drive to achieve the given illuminance output.

[0211] The illuminance in the garage is related to many factors such as the parameters of the lamps themselves, the type of the garage floor, the garage walls, and the service life of the garage, and it is a non-linear complex system.

[0212] As Figure 9 and Figure 10 shown, the present invention proposes to use RBF neural network PID control to achieve illuminance adjustment, which solves the non-linearity, unknown model parameters, and achieves better control.

[0213] 1. Structure of RBF neural network:

[0214] Assume that the network is a three-layer feedforward structure:

[0215] Input layer (2 nodes):

[0216]

[0217] Hidden layer (m nodes): Gaussian radial basis function

[0218]

[0219] Where:

[0220] The center vector of the j-th hidden node;

[0221] σ j : The width of the j-th hidden node;

[0222] Output layer (3 nodes): PID parameters

[0223]

[0224] Where w pj , w ij, w dj is the output weight, b p , b i , b d is the bias term; 2. PID control law:

[0225] Dynamic dimming output:

[0226]

[0227] u(t) is the controller output;

[0228] e(t) is the error signal (the difference between the set value and the actual value);

[0229] Kp, Ki, and Kd are the proportional, integral, and derivative gains respectively.

[0230] 3. Parameter learning algorithm derivation

[0231] 1). Objective function definition

[0232] Define the loss function:

[0233]

[0234] Where:

[0235] W = [w pj , w ij , w dj : Weight matrix

[0236] C = [C j : Set of center vectors

[0237] ∑ = [σ j Width vector

[0238] λ: Regularization coefficient

[0239] 2). Gradient descent update rule

[0240] Weight update:

[0241]

[0242] Where m ∈ {p, i, d}, the partial derivative is:

[0243]

[0244] Center vector update:

[0245]

[0246] Where:

[0247]

[0248] Width update:

[0249]

[0250] Where:

[0251]

[0252] 4. Stability proof (Lyapunov method)

[0253] 1). Construct the Lyapunov function

[0254] Where is the parameter error.

[0255] 2). Analysis of the time derivative

[0256] Substituting the update rule, we get:

[0257]

[0258] It is proved that the system is globally asymptotically stable.

[0259] Program simulation and debugging:

[0260] To verify the rationality and effectiveness of the algorithm of the present invention, simulation verification of fuzzy control and RBF neural network PID control was carried out using matlab.

[0261] In the first step, two inputs were given to verify the performance of the fuzzy controller.

[0262] In the second step, a curve (simulating the illuminance reference curve) was given to verify the dynamic performance of the RBF neural network PID control.

[0263] Performance of the fuzzy controller:

[0264] 1) Two inputs were given as the traffic flow and the illuminance in the intersection area;

[0265] 2) Design of the membership function: Triangular or Gaussian functions were adopted to cover the entire input range.

[0266] 3) Fuzzy rules

[0267] Construction of the rule base: "IF-THEN" rules were defined based on expert experience, for example:

[0268] IF Q is small and E is small, then u is S.

[0269] Such as Figure 11As shown, the transition plane of the fuzzy controller rules is presented. The results show that the established fuzzy rules have a smooth transition, avoiding jumps in the control quantity, and the design is reasonable.

[0270] 4) Inference method: The Mamdani min-max inference method is adopted to calculate the activation strength of each rule.

[0271] 5) Defuzzification

[0272] Output membership function: The triangular distribution is adopted to cover the control quantity range.

[0273] Defuzzification method: The center of gravity method (COG) is selected to generate the precise control quantity.

[0274] For the given two hypothetical data, as Figure 12 shown, the weighted calculation process of the output membership degree is presented.

[0275] 6) Performance evaluation of the fuzzy controller

[0276] 1. Overlap rate of membership functions: It is set to 25% - 30% to ensure a smooth transition and avoid jumps in the control quantity.

[0277] 2. Rule weight assignment: Adjust the rule weights according to the experimental data. For example, the rule weight of "high traffic flow and high illuminance" is increased by 20% to optimize the response speed.

[0278] 3. Defuzzification accuracy: By adjusting the resolution of the output membership degree, the quantization error of the control quantity is controlled within ±1%.

[0279] Performance analysis of RBF neural network PID control:

[0280] Given a sine wave as the illuminance reference curve, the RBF neural network of the present invention is simulated and verified. The verification results are as Figure 13 shown. The red line is the given reference, and the black dotted line is the controller tracking.

[0281] Through simulation analysis, the RBF neural network PID control has the following excellent performances:

[0282] 1) Optimization of dynamic response

[0283] For the given reference curve, the RBF neural network PID can perform fast tracking, and the dynamic error is within ±1.

[0284] Adaptability:

[0285] At t = 0.5s at the peak (or trough) position of the sine wave, the controller can track well, indicating good adaptability to sudden changes in the reference.

[0286] 3) Strong robustness

[0287] It has strong anti-interference ability, and the steady-state error is maintained within ±2.

[0288] The simulation results show that the RBF neural network PID control designed by the present invention exhibits excellent non-linear adaptability, dynamic response speed and anti-interference ability in the dynamic dimming scenario of the underground garage. Its algorithm design is reasonable and meets the engineering requirements.

[0289] The present invention also proposes a computer-readable storage medium storing a computer program, and when the program is executed by a processor, the underground garage lighting control method is implemented.

[0290] The present invention has the following innovation points:

[0291] 1. After studying the mechanism of human eye dark adaptation, starting from the light and dark adaptation time and the law of brightness change, a comfort illuminance model for the garage is established, and the optimal illuminance reference for the garage is given. This model fully considers comfort and human eye adaptability, and at the same time considers the influence of factors such as seasonal changes, weather changes, and traffic flow on illuminance, achieving lighting on demand and comfortable lighting. When the traffic flow is large and the brightness difference between inside and outside the garage is large and the illuminance needs to be increased, a high illuminance is output. When the traffic flow is small, there are few people passing by, and at night, a low illuminance is output. While ensuring comfort and safety, the power consumption of the garage is greatly reduced.

[0292] 2. The working mechanism of the garage intersection area is studied, and a fuzzy controller with traffic flow and basic garage illuminance as inputs and intersection area illuminance reference as output is established. Fuzzy control rules are established with the help of expert experience, and complex control is simplified and made efficient through fuzzy reasoning. The established fuzzy controller does not require complex training and is suitable for illuminance adjustment control with high real-time requirements. By dynamically adjusting the illuminance of the intersection area in real time, on the one hand, the operation safety of the garage is improved, and on the other hand, the power consumption is reduced.

[0293] 3. According to the optimal illuminance curve of the garage generated by the garage comfort illuminance model and the intersection area fuzzy control, the RBF artificial neural network PID algorithm is used to adjust the illuminance of the lamps in each area in real time to achieve closed-loop tracking control. The artificial neural network RBF-PID control in this algorithm has the following control advantages, thus realizing a more intelligent lighting control

[0294] (1) Strong non-linear adaptability: Approximating complex mappings through Gaussian kernel functions to solve the dynamic coupling problem between traffic flow and illuminance

[0295] (2) Online self-learning optimization: Online adjusting the PID parameters to adapt to the long-term dynamic effects of differences in different types of garages, seasonal changes, equipment aging, etc., and giving the most suitable illuminance control for this garage.

[0296] (3) Strong anti-interference ability: It has better robustness than traditional PID in low light or sensor noise environments.

[0297] 4. Eliminate objective interference factors. The illuminance in the garage is directly related to the wall reflectance, floor reflectance, cleanliness of the garage, cleanliness of the lamps, lamp aging, etc. Traditional dimming control cannot consider these objective factors. This algorithm self-learns and optimizes, uses the neural network RBF to adjust the PID parameters online, adapts to the long-term dynamic impacts of different types of garages, seasonal changes, equipment aging, etc., and gives the optimal control.

[0298] Through the description of the above implementation manners, those skilled in the art can clearly understand that each implementation manner can be realized by means of software plus a necessary general hardware platform, and of course, it can also be realized by hardware. Based on such an understanding, the essence of the above technical solution, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disc, etc., and includes several instructions to enable a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0299] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions described in the foregoing embodiments, or perform equivalent replacements for some of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. An underground garage lighting control system based on artificial neural network PID divides the underground garage into an entrance section, a transition section, a basic section, and an intersection area, and is characterized in that Including: The detection unit includes brightness sensors, traffic flow sensors, illuminance sensors, temperature and humidity sensors, and microwave speed sensors arranged inside and outside the garage. The intelligent connected lamp includes an intelligent gateway unit, an LED dimming drive unit, and an LED lamp. The intelligent gateway unit is used to receive control instructions and communicate with the upper computer. The upper computer is used to receive the data of the detection unit, generate the optimal illuminance control curve based on the garage comfort illuminance model and the intersection area fuzzy control model, and adjust the illuminance of each area lamp in real time through the RBF neural network PID algorithm.

2. The underground garage lighting control system based on artificial neural network PID according to claim 1, wherein One brightness sensor is arranged inside and outside the garage entrance and exit. The illuminance sensors are arranged at the garage entrance, garage ramp, middle of the garage, and garage intersection area. One temperature and humidity sensor is arranged inside and outside the garage. One traffic flow sensor is arranged in each of the four directions in the intersection area. One microwave speed sensor is arranged in each of the four directions in the intersection area.

3. A method for controlling the lighting of an underground garage using the system according to claim 1 or 2, characterized in that, Including the following method steps: Collect the traffic flow in the intersection area, the brightness inside and outside the garage, and the illuminance of the entrance section, transition section, basic section, and intersection area inside the garage. Calculate the illuminance reference values of the entrance section, transition section, and basic section according to the garage comfort illuminance model. Combined with the current traffic flow, give the illuminance reference value of the intersection area according to fuzzy control. Draw the optimal illuminance curve of the garage. Use the given optimal illuminance curve and use the RBF neural network PID to control the lamp drive until the optimal illuminance is achieved.

4. The method according to claim 3, wherein The garage comfort illuminance model includes algorithm models corresponding to the entrance section, transition section, and basic section, where: Algorithm model of the entrance section: The entrance section is set in the entrance area of the garage, and the illuminance reference of the entrance section is defined as follows: L out Outside the garage; brightness; k reduction factor, with a value of 0.8; Algorithm model of the transition section: The transition section is set in the ramp area where the vehicle enters the garage, and the illuminance reference of the transition section is defined as follows: L in Brightness inside the garage; L out Brightness outside the garage; X Distance of the vehicle from the starting point of the transition section; L total Length of the transition section Algorithm model of the basic section: The basic section is located inside the garage. Taking the traffic flow Q as a variable, a dynamic illuminance adjustment formula is proposed: E b : Basic illuminance, taking 50 LX for public garages and 30 LX for residential garages according to national standards; Q: Measured vehicle flow, the number of vehicles passing through per unit time; Q max : Maximum vehicle flow; α: Adjustment coefficient; Calibration method of the adjustment coefficient α: Optimize the α value in real time by combining the Kalman filter: α t = α t-1 + K(Q n - Q n-1 ); Where K is the Kalman gain coefficient; Nonlinear relationship modeling: When the traffic flow exceeds the threshold, the illuminance demand increases exponentially. A piecewise function is introduced:

5. The method according to claim 3, characterized in that The rule base of the fuzzy control includes the following rules: If the traffic flow is large and the illuminance of the basic section is high, the illuminance of the intersection area is set to high. If the traffic flow is large and the illuminance of the basic section is low, the illuminance of the intersection area is set to high. If the traffic flow is small and the illuminance of the basic section is low, the illuminance of the intersection area is set to low.

6. The method according to claim 3, wherein The reasoning of the fuzzy control includes the following steps: Fuzzification: The traffic flow detection value Q(K) and the illuminance E(K) of the basic section are used as the inputs of the fuzzy controller, and the output of the fuzzy controller is the illuminance value u of the intersection area. Define the subsets of the traffic flow Q(K), the illuminance E(K) of the basic section, and the illuminance reference value u of the intersection area as {small, medium, large}, and the universe of discourse is [0,1]; The triangular membership function is adopted. Fuzzy reasoning: The fuzzy reasoning adopts Mamdani reasoning, and the reasoning algorithm is as follows: η i = μ ei [e(k)] ∧ μ eci [ec(k)] = min(μ ei [e(k)] · μ eci [ec(k)]); η i is the fuzzy inference calculated value; μ ei [e(k)] is the traffic flow membership value; μ eci [ec(k)] is the basic section illuminance membership value; ∧ is the minimum operation; fuzzy rule base establishment; Defuzzification: The defuzzification adopts the centroid method; The defuzzification calculation method is as follows:

7. The method according to claim 3, wherein The RBF neural network PID controls the lamp drive Including the following steps: Receive the optimal illuminance curve; Collect the current illuminance and calculate the error; The illuminance of the lamps in each area is adjusted in real time through the RBF artificial neural network PID algorithm.

8. The method according to claim 7, wherein The RBF artificial neural network PID algorithm includes: The RBF neural network structure, and the network is set as a three-layer feedforward structure: Input layer (2 nodes): Hidden layer (m nodes): Gaussian radial basis function: Where: The center vector of the j-th hidden node; σ j : The width of the j-th hidden node; Output layer (3 nodes): PID parameters: where w pj , w ij , w dj is the output weight, b p , b i , b d is the bias term; PID control law, dynamic dimming output: u(t) is the output of the controller; e(t) is the error signal (the difference between the set value and the actual value); Kp, Ki, and Kd are the proportional, integral, and differential gains respectively.

9. A computer-readable storage medium storing a computer program, characterized in that, When the program is executed by the processor, it implements the underground garage lighting control method described in any one of claims 3-8.

Images