def main():
# 读取用电负荷数据集
length_load, time_load, load = load_data.read_load_history(is_full=True)
time_load = time_load.reshape(length_load, 4)
time_load = time_load[:, 1:]
load = load.reshape(length_load, 20)
# 对每个区域用电负荷进行预处理与预测
for zone_id in range(18, 21):
zone_load_series = load[:, zone_id - 1]
clustering(zone_id, time_load, zone_load_series)
def fill_missing_data():
# 获取历史用电负荷数据
length_load, time_load, load = load_data.read_load_history(is_full=True)
# 获取历史温度数据
length_temperature, time_temperature, temperature = load_data.read_temperature_history(
)
print('Filling missing data...')
time_temperature = time_temperature.reshape((len(time_temperature), 4))
temperature = temperature.reshape((len(temperature), 1))
load = load.reshape((len(load), 1))
# 分别记录填充缺失值后的时间戳、温度和用电负荷数据
time_temperature_full = time_temperature[0:39414]
print('00:', len(time_temperature_full), time_temperature_full[-1])
temperature_full = temperature[0:39414]
load_full = load[0:39414]
step_size = 365 * 24
for i in range(168 + 18):
index = len(temperature_full)
time = np.zeros((1, 4))
print('xx:', index - step_size - 24)
time = np.add(time, time_temperature_full[index - step_size - 24])
time[0][0] += 1
time_temperature_full = np.append(time_temperature_full, time, axis=0)
load_tmp = np.zeros((1, 1))
load_tmp = np.add(load_tmp, load_full[index - 7 * 24])
load_full = np.append(load_full, load_tmp, axis=0)
temperature_tmp = np.zeros((1, 1))
for j in range(1, 2):
prev_index = index - j * step_size - 24
prev_temp = temperature_full[prev_index]
temperature_tmp = np.add(temperature_tmp, prev_temp)
temperature_tmp = temperature_tmp / 4
temperature_full = np.append(temperature_full, temperature_tmp, axis=0)
# 写入csv文件
df_time = pd.DataFrame(time_temperature_full)
df_time.to_csv(r'./LSTM/dataset/time_full.csv', header=False, index=False)
df_temperature = pd.DataFrame(temperature_full)
df_temperature.to_csv(r'./LSTM/dataset/temperature_full.csv',
header=False,
index=False)
df_load = pd.DataFrame(load_full)
df_load.to_csv(r'./LSTM/dataset/load_full.csv', header=False, index=False)
print('Done')
def set_training_data():
# 读取数据集
length_temperature, time_temperature, temperature = load_data.read_temperature_history(
is_full=True)
length_load, time_load, load = load_data.read_load_history(is_full=True)
# 归一化
time_temperature = time_temperature.reshape(len(time_temperature), 4)
mean_time, std_time, time_temperature = normalization(time_temperature)
temperature = temperature.reshape(len(temperature), 11)
mean, std, temperature = normalization(temperature)
load = load.reshape(len(load), 20)
mean_load, std_load, load = normalization(load, if_log=True)
time_temperature = time_temperature.reshape((len(time_temperature), 1, 4))
temperature = temperature.reshape((len(temperature), 1, 11))
load = load.reshape((len(load), 1, 20))
# 使用以前的用电负荷数据作为预测的输入
load_for_inputs = np.copy(load)
for i in range(len(load_for_inputs) - 1, -1, -1):
index = i - 24 * 7
if index < 0:
index = i
load_for_inputs[i] = np.copy(load_for_inputs[index])
inputs_full = np.concatenate(
(temperature, time_temperature, load_for_inputs), axis=2)
inputs_full = inputs_full.reshape((len(inputs_full), 1, 35))
load = load.reshape(len(load), 20)
# inputs_full = inputs_full[0:39414]
inputs = inputs_full[0:len(load) - 24 * 7]
# outputs_full = load[0:39414]
outputs_full = load
outputs = load[0:len(load) - 24 * 7]
print "Full inputs shape:", inputs_full.shape
print "Full outputs shape:", outputs_full.shape
print "Input shape:", inputs.shape
print "Output shape:", outputs.shape
return mean_load, std_load, inputs_full, inputs, outputs_full, outputs
y_1, = plt.plot(expected_output, color='blue')
y_2, = plt.plot(predicted_output, color='green')
legend_var.append(y_1)
legend_var.append(y_2)
plt.title(plot_name + str(zone_id), fontproperties=font_title)
plt.legend(legend_var, legend_name, prop=font)
plt.ylabel(u"用电量/kWh", fontproperties=font)
plt.xlabel(u"时间/h", fontproperties=font)
plt.show()
def plot_feature(data_frame):
plt.figure()
data_frame.plot(kind='barh', x='feature', y='fscore')
font = FontProperties(fname='C:\Windows\Fonts\simsun.ttc', size=12)
font_title = FontProperties(fname='C:\Windows\Fonts\simsun.ttc', size=14)
plt.title(u"XGBoost模型:特征重要性排名", fontproperties=font_title)
plt.xlabel(u"相对重要性", fontproperties=font)
plt.show()
if __name__ == "__main__":
# 绘制原始数据曲线图
length_load, time_load, load_load = load_data.read_load_history()
plot_origin(length_load, time_load, load_load, 20, u"用电负荷历史数据：zone-",
u"用电量/kWh")
length_temp, time_temp, temp_temp = load_data.read_temperature_history()
plot_origin(length_temp, time_temp, temp_temp, 11, u"历史温度数据：station-",
u"温度/℃")