DIY Rain Prediction Using Arduino, Python and Keras

1_id has_rain  humidity_10percentile  humidity_30percentile  humidity_70percentile  humidity_90percentile          ...           temperature_steady_4  temperature_steady_5  temperature_steady_6  temperature_steady_7  temperature_steady_8                                                                                                                         ...                                                                                                                       
204_27_2017_3         0                   36.0                   44.8                   61.0                   63.0          ...                            NaN                   NaN                   NaN                   NaN                   NaN
304_28_2017_0         0                   68.0                   70.0                   74.0                   75.0          ...                           14.0                   NaN                   NaN                   NaN                   NaN
404_28_2017_1         0                   40.0                   45.0                   63.2                   69.0          ...                           20.0                  14.0                   NaN                   NaN                   NaN
504_28_2017_2         0                   34.0                   35.9                   40.0                   41.0          ...                           29.0                  20.0                  14.0                   NaN                   NaN
604_28_2017_3         0                   36.1                   40.6                   52.0                   54.0          ...                           19.0                  29.0                  20.0                  14.0                   NaN
704_29_2017_0         0                   52.0                   54.0                   56.0                   58.0          ...                           26.0                  19.0                  29.0                  20.0                  14.0
804_29_2017_1         0                   39.4                   43.2                   54.6                   57.0          ...                           18.0                  26.0                  19.0                  29.0                  20.0
904_29_2017_2         1                   41.0                   42.0                   44.2                   47.0          ...                           28.0                  18.0                  26.0                  19.0                  29.0 

1_id date  humidity  light  pressure  rain  temperature
2                                                                            
304_27_2017_03_08   2017-04-27 03:08:36      67.0    0.0    1007.0   0.0         11.0
404_27_2017_03_19   2017-04-27 03:19:05      66.0    0.0    1007.0   0.0         11.0
504_27_2017_03_29   2017-04-27 03:29:34      66.0    0.0    1007.0   0.0         11.0 

1conda activate weather
2
3python serial_listener.py
4
5# every  10 minutes you should see data from the weather station coming in :
6
7[Sensor:  type(temperature), value(14.3)]
8
9[Sensor: type(pressure), value(1056.0)]
10
11...

1from sklearn.preprocessing import StandardScaler
2
3from sklearn.externals  import joblib
4
5..
6
7scaler = StandardScaler()
8
9X = scaler.fit_transform(X)  
10
11...
12
13# of course we should be careful to save the scaled model for  later reuse
14
15joblib.dump(scaler, 'model_file_name.save') 

1cd weather-predict 
2# create anaconda environment named "weather"  with python 3.6.2
3conda create --name weather python=3.6.2 
4
5# activate  environment
6conda activate weather
7
8# install all packages 
9pip install  -r requirements.txt 

1from keras.models import Sequential
2
3from keras.layers import Dense
4from  keras.layers import Dropout 
5...
6input_dimensions = X.shape[1] 
7optimizer  = 'rmsprop'
8dropout = 0.05
9model = Sequential()
10inner_nodes = int(input_dimensions  / 2)
11model.add(Dense(inner_nodes, kernel_initializer='uniform', activation='relu',  input_dim=input_dimensions))
12model.add(Dropout(rate=dropout))
13model.add(Dense(inner_nodes,  kernel_initializer='uniform', activation='relu'))
14model.add(Dropout(rate=dropout))
15model.add(Dense(1,  kernel_initializer='uniform', activation='sigmoid'))
16model.compile(optimizer=optimizer,  loss='mean_absolute_error', metrics=['accuracy']) 
17
18model.fit(X, y, batch_size=1,  epochs=50)
19...
20
21# save the model for later use
22classifier.save('file_model_name')

1{	
2        "_id" : "04_27_2017_06_17",
3	"humidity" : 65,
4	"date"   : ISODate("2017-04-27T06:17:18Z"),
5	"pressure" : 1007,
6	"temperature"   : 9,
7	"rain" : 0,
8	"light" : 15
9}

1from sklearn.model_selection import train_test_split 
2
3...
4
5main_data,   test_data = train_test_split(dataframe, test_size=percent_test_data) 
6
7...

1mongodb = {
2    'host': 'localhost',
3    'port': 27017
4}
5...

1client = MongoClient(host, port).weather.datapoints 
2
3cursor = client.find(
4            {'$and' : [
5                {'date' : {'$gte' : start_date}},
6                {'date' : {'$lte' : end_date}}
7            ]}
8        )
9data  = list(cursor)
10..

1client = MongoClient(host, port).weather.datapoints 
2
3cursor = client.find(
4
5            {'$and' : [
6                {'date' : {'$gte' : start_date}},
7
8                {'date' : {'$lte' : end_date}}
9            ]}
10        )
11data
12  = list(cursor)
13..

1#include "LowPower.h"<br>#include "SoftwareSerial.h"
2#include "Wire.h"
3#include  "Adafruit_Sensor.h"
4#include "Adafruit_BME280.h"
5#include "BH1750.h"
6
7SoftwareSerial  serialComm(4, 5); // RX, TX
8Adafruit_BME280 bme; 
9BH1750 lightMeter;
10const  byte rainPin = A0;
11
12byte sensorsCode = 1;
13/**
14 * voltage level that  will pun the microcontroller in deep sleep instead of regular sleep
15 */
16int  voltageDeepSleepThreshold = 4200; 
17const byte peripherialsPowerPin = 6;
18char  buffer[] = {' ',' ',' ',' ',' ',' ',' '};
19
20struct sensorData
21  {
22      byte  humidity;    
23      int temperature;
24      byte rain;
25      int  pressure;
26      long voltage;
27      int light;
28  };
29
30sensorData sensors;
31
32void  setup() 
33{  
34    Serial.begin(9600);
35    serialComm.begin(9600);
36    pinMode(peripherialsPowerPin,  OUTPUT);
37    digitalWrite(peripherialsPowerPin, HIGH);
38    delay(500);
39    if (!bme.begin()) {
40        Serial.println("Could not find a valid BME280  sensor, check wiring!");
41        while (1) {
42           customSleep(100);
43        }
44    }
45    Serial.println("Initialization finished succesfully");
46    delay(50);
47    digitalWrite(peripherialsPowerPin, HIGH);
48}
49
50void  loop() 
51{    
52    updateSenzors();
53    transmitData();    
54    customSleep(75);       
55}
56
57void updateSenzors() 
58{
59    bme.begin();
60    lightMeter.begin();
61    delay(300);
62    sensors.temperature = bme.readTemperature();
63    sensors.pressure  = bme.readPressure() / 100.0F;
64    sensors.humidity = bme.readHumidity();
65    sensors.light = lightMeter.readLightLevel();
66    sensors.voltage = readVcc();
67    sensors.rain = readRain();
68}
69
70void transmitData()
71{
72    emptyIncommingSerialBuffer();
73    Serial.print("Temp:");Serial.println(sensors.temperature);
74    Serial.print("Humid:");Serial.println(sensors.humidity);
75    Serial.print("Pressure:");Serial.println(sensors.pressure);
76    Serial.print("Light:");Serial.println(sensors.light);
77    Serial.print("Voltage:");Serial.println(sensors.voltage);
78    Serial.print("Rain:");Serial.println(sensors.rain);
79    transmitSenzorData("T", sensors.temperature);
80    transmitSenzorData("H",  sensors.humidity);
81    transmitSenzorData("PS", sensors.pressure);
82    transmitSenzorData("L",  sensors.light);
83    transmitSenzorData("V", sensors.voltage);
84    transmitSenzorData("R",  sensors.rain);
85}
86
87void emptyIncommingSerialBuffer()
88{
89    while (serialComm.available()  > 0) {
90        serialComm.read();
91        delay(5);
92   }
93}
94
95void  transmitSenzorData(String type, int value)
96{
97    serialComm.print(type);
98    serialComm.print(sensorsCode);
99    serialComm.print(":");
100    serialComm.print(value);
101    serialComm.print("|");
102    delay(50);
103}
104
105void customSleep(long  eightSecondCycles)
106{
107    if (sensors.voltage > voltageDeepSleepThreshold)  {
108        delay(eightSecondCycles * 8000);
109        return;
110    }
111    digitalWrite(peripherialsPowerPin,  LOW);
112    for (int i = 0; i < eightSecondCycles; i++) {
113        LowPower.powerDown(SLEEP_8S,  ADC_OFF, BOD_OFF);  
114    }
115    digitalWrite(peripherialsPowerPin, HIGH);
116    delay(500);
117}
118
119byte readRain()
120{
121    byte level = analogRead(rainPin);
122
123    return map(level, 0, 1023, 0, 100);  
124}
125
126long readVcc() {
127    //  Read 1.1V reference against AVcc
128    // set the reference to Vcc and the measurement  to the internal 1.1V reference
129    #if defined(__AVR_ATmega32U4__) || defined(__AVR_ATmega1280__)  || defined(__AVR_ATmega2560__)
130      ADMUX = _BV(REFS0) | _BV(MUX4) | _BV(MUX3)  | _BV(MUX2) | _BV(MUX1);
131    #elif defined (__AVR_ATtiny24__) || defined(__AVR_ATtiny44__)  || defined(__AVR_ATtiny84__)
132      ADMUX = _BV(MUX5) | _BV(MUX0);
133    #elif  defined (__AVR_ATtiny25__) || defined(__AVR_ATtiny45__) || defined(__AVR_ATtiny85__)
134      ADMUX = _BV(MUX3) | _BV(MUX2);
135    #else
136      ADMUX = _BV(REFS0) |  _BV(MUX3) | _BV(MUX2) | _BV(MUX1);
137    #endif  
138  
139    delay(2); // Wait  for Vref to settle
140    ADCSRA |= _BV(ADSC); // Start conversion
141    while  (bit_is_set(ADCSRA,ADSC)); // measuring
142    uint8_t low  = ADCL; // must read  ADCL first - it then locks ADCH  
143    uint8_t high = ADCH; // unlocks both
144    long result = (high<<8) | low;
145    result = 1125300L / result; // Calculate  Vcc (in mV); 1125300 = 1.1*1023*1000
146    
147    return result; // Vcc in millivolts
148}

1cd weather-predict 
2# create anaconda environment named "weather"
3  with python 3.6.2
4conda create --name weather python=3.6.2 
5
6# activate
7  environment
8conda activate weather
9
10# install all packages 
11pip install
12  -r requirements.txt 

1from sklearn.preprocessing import StandardScaler
2
3from sklearn.externals
4  import joblib
5
6..
7
8scaler = StandardScaler()
9
10X = scaler.fit_transform(X)
11  
12
13...
14
15# of course we should be careful to save the scaled model for
16  later reuse
17
18joblib.dump(scaler, 'model_file_name.save') 

1_id has_rain  humidity_10percentile  humidity_30percentile  humidity_70percentile
2  humidity_90percentile          ...           temperature_steady_4  temperature_steady_5
3  temperature_steady_6  temperature_steady_7  temperature_steady_8                                                                                                                         ...
4                                                                                                                       
504_27_2017_3
6         0                   36.0                   44.8                   61.0
7                   63.0          ...                            NaN                   NaN
8                   NaN                   NaN                   NaN
904_28_2017_0
10         0                   68.0                   70.0                   74.0
11                   75.0          ...                           14.0                   NaN
12                   NaN                   NaN                   NaN
1304_28_2017_1
14         0                   40.0                   45.0                   63.2
15                   69.0          ...                           20.0                  14.0
16                   NaN                   NaN                   NaN
1704_28_2017_2
18         0                   34.0                   35.9                   40.0
19                   41.0          ...                           29.0                  20.0
20                  14.0                   NaN                   NaN
2104_28_2017_3
22         0                   36.1                   40.6                   52.0
23                   54.0          ...                           19.0                  29.0
24                  20.0                  14.0                   NaN
2504_29_2017_0
26         0                   52.0                   54.0                   56.0
27                   58.0          ...                           26.0                  19.0
28                  29.0                  20.0                  14.0
2904_29_2017_1
30         0                   39.4                   43.2                   54.6
31                   57.0          ...                           18.0                  26.0
32                  19.0                  29.0                  20.0
3304_29_2017_2
34         1                   41.0                   42.0                   44.2
35                   47.0          ...                           28.0                  18.0
36                  26.0                  19.0                  29.0 

1_id date  humidity  light  pressure  rain  temperature
2                                                                            
304_27_2017_03_08  2017-04-27 03:08:36      67.0    0.0    1007.0   0.0         11.0
404_27_2017_03_19  2017-04-27 03:19:05      66.0    0.0    1007.0   0.0         11.0
504_27_2017_03_29  2017-04-27 03:29:34      66.0    0.0    1007.0   0.0         11.0 

1from keras.models import Sequential
2
3from keras.layers import Dense
4from
5  keras.layers import Dropout 
6...
7input_dimensions = X.shape[1] 
8optimizer
9  = 'rmsprop'
10dropout = 0.05
11model = Sequential()
12inner_nodes = int(input_dimensions
13  / 2)
14model.add(Dense(inner_nodes, kernel_initializer='uniform', activation='relu',
15  input_dim=input_dimensions))
16model.add(Dropout(rate=dropout))
17model.add(Dense(inner_nodes,
18  kernel_initializer='uniform', activation='relu'))
19model.add(Dropout(rate=dropout))
20model.add(Dense(1,
21  kernel_initializer='uniform', activation='sigmoid'))
22model.compile(optimizer=optimizer,
23  loss='mean_absolute_error', metrics=['accuracy']) 
24
25model.fit(X, y, batch_size=1,
26  epochs=50)
27...
28
29# save the model for later use
30classifier.save('file_model_name')

1conda activate weather
2
3python serial_listener.py
4
5# every
6  10 minutes you should see data from the weather station coming in :
7
8[Sensor:
9  type(temperature), value(14.3)]
10
11[Sensor: type(pressure), value(1056.0)]
12
13...

1from sklearn.model_selection import train_test_split 
2
3...
4
5main_data,  test_data = train_test_split(dataframe, test_size=percent_test_data) 
6
7...

1mongodb = {
2    'host': 'localhost',
3    'port': 27017
4}
5...

1{	
2        "_id" : "04_27_2017_06_17",
3	"humidity" : 65,
4	"date"  : ISODate("2017-04-27T06:17:18Z"),
5	"pressure" : 1007,
6	"temperature"  : 9,
7	"rain" : 0,
8	"light" : 15
9}